Brainstorming ideas for new football sim [Archive]

sabotai

05-30-2003, 11:23 PM

Ok, so I have sat down and started just writing things down that I want to include in my next football sim (the one that will be done right).

I settled on a few things.

1) Single-player and single-player only. I know there are people out there that want multi-player, but based on what I have seen, multi-player text sim is a niche market within a niche market. IMO, when you make a game have multi-player capabilities, you are forced to program multi-player first / single-player second if onyl because you want to make it easy for league commishes. If you programmed single-player first / multi-player second, I think you end up with things that would confuse and frustrate league commishes and then they just don't use your product.

So I have settled on that my next football sim attempt will be single-player.

2) Stats, and lots of them. This goes without saying. And now that I have settled on no multi-player, it allows me to not worry too much about disk space (no transfering files over internet). Which means I'll be able to do much more in way of stats and record keeping.

3) A realistic model of the salary cap and free agency. We all know that in FOF, there are no transition players and that franchise players and RFA's do not work in FOF the way they do in real life. THere are also no contract incentives and bonuses. What I want to do is have a true to life representation of the NFL free agency, salary cap and contracts.

4) Varying league sizes. This feature was in FLS, but it'll be more limited in my next game, is only for the sake of sheduling functions. I'm thinking along the lines of league sizes from 24-32 teams, maybe up to 36. We'll see. But it will not be 8-48 like last time. Mainly for 3 reasons. 1) Sheculing functions take time to program right. 2) Talent distribution gets too thick or thin 3) AI anomolies at low number teams.

Not really anomolies, just things you don't want to happen. Like say the best QB in the league goes on FA. No way he goes unsigned. But there are only 8 teams. It's very possible none of the 8 teams would want to or could cut their starting QB or that they simply don't have the cap room.

5) Attributes. I'm leaning more towards not representing attributes. I'll keep them all hidden. The way you find out about a player is 1) Watch them. 2) YOur scouts and coaches clue you into things like "He's not the best athlete, but he plays with a lot of intensity and determination." Obviously, you'll get a lot more input than one sentance, but that's pretty much the idea. 3) Drill and work-out stats. Like how fast he runs the 40 and how many reps of X weight he can do.

One worry I have about this though is getting into the game. When you start off a game, you get thrown in blind. I'm worried that by doing it this way, it'll be harder for people to jump into the game without just being able to go "Speed 95 Catch 90" He's my starting receiver.

In short, my worry is that people won't get into it quickly and thus not play it. Thoughts?

6) Plays, play designer, game plans.

What about your thoughts? Any ideas? Basically, what I'm looking for are ideas that may not work well for multi-player, but will work great with single-player. So if anyone has any general ideas they have for me, PLEASE let me know. I'm begging here! ;)

tucker342

05-30-2003, 11:39 PM

maybe you should have an option so that people can have the attributes showing, or people can have it hidden. Kind of like how they have it in OOTP.

Sounds promising:)

EagleFan

05-30-2003, 11:40 PM

I like the idea of somewhat hidden attributes. Just keep in mind with the work-out stats that the best football players aren't always the best work-out numbers and the other way around (Mike Mamula comes to mind, oh the painful memories of a wasted pick--unless you're talking about the FOF universe where he seemed to be quite good).

The hidden attributes will cause a problem if you are allowing leagues to start with an initial draft. That is where the problems of not really knowing anything at all about a player will cause problems balancing out an AI that will be drafting the other teams.

Play design will open the game up to problems of a "super play". Depending on the extent of play design. That isn't actually something that is a big problem as most who really want to play it wouldn't be looking for a loop-hole like that but it would become a source of criticism by some.

I've been working on a text sim for a different sport in my spare time and scrapped the play designer when I spent too much time on it because I kept finding holes that it would open up, which would mean more checks and balances every time I found a new problem. It was one big bottomless pit, or so it felt.
If I ever get it close enough to be playable I'll post it on here for everyone as it's currently just a hobby and something to get a feel for sports programming.

sabotai

05-30-2003, 11:58 PM

tucker, I think that makes the most the sense. I'll allow them to have them showing for those who want to have them showing. I'll have to work this out. Maybe I'll have it so that you'll see several overal attributes, but not the specific ones....we'll see.

Eagles, good point with the initial draft. And as for work-out stats, these just show a guy's athletic ability, and do not translate as great work-out = great player, as you said. Work-out stats will only show a player's athletic ability. Most of the attributes for a player are mental-like attributes. Like intensity, intellegence, instinct, etc.

Play-design. This will be limited. In FBPro, it allowed you to draw your own routes and running lanes. This was part of the problem. In my game, you'll have to stick with routes, running lanes, blocking schemes, etc. that are programmed into it. That way I can cut down on the possible "AI buster". Which was the second problem in FBPro. The defense AI was broken. I think that was the biggest problem with FBPro (and basically every football game ever made) that allowed for super plays.

One other thing to keep in mind, (I plan to have this) that once a play that you designed is used in a game, all the other teams get access to that play. IOW, it'll be impossible for you to have plays and just run those that the AI will not be able to use themselves.

If it does get too out of hand, and I hope it doesn't, I'll just do as you did and scrap the play designer and just make tons o' stock plays.

EagleFan

05-31-2003, 12:04 AM

I like your idea of giving the play to all the teams to use after you have used it. That is the essence of the NFL at this point. See something that works and copy it.

Not to over-complicate things, which I tend to do :), but any plans on the counter-play idea. You create a play that you use and eventually a counter play gets created by another team for it (like a defense to stop it). Sorry, now you can see how I led myself down that endless path when I started doing the play designing.

sabotai

05-31-2003, 12:19 AM

"Not to over-complicate things, which I tend to do , but any plans on the counter-play idea. You create a play that you use and eventually a counter play gets created by another team for it (like a defense to stop it). Sorry, now you can see how I led myself down that endless path when I started doing the play designing."

I HOPE, I really hope, and I hope some more, that I will be able to create an AI that will be able to create it's own plays, both offensive and defensive.

When you think about it, play creation in the NFL is not all that complicated. You have 5 linement who block, you have 1 QB who takes the snap who can either pass or run, and you have 4 backs who can receive, pass or throw.

On defense, it's even less complicated.

So basically, to answer your question, yes, I hope that I can create an AI that can come up with specific defensive schemes to stop specific offensive moves. Spotting and predicting the offensive moves is a different story. :)

The more I read about offensive and defensive logic, schemes, assignments, etc. the less complicated it seems. A lot of what you do one defensie is basically 1) Do initial assignment 2) Determine direction and type of play 3) Respond accordingly 4) Repeat steps 2 and 3 until end of play. In a simplified explaination, that's the defense. So really with defensive play design, it all starts and ends with step 1. So if a team likes to attack a specific spot, the AI should pick up on this a design specific defense plays to attack and defend that spot.

With offensive play design, it all comes down to what your strengths are, and how to best use them against the weaknesses of your opponents. It's really not all that complicated.

Predicting and spotting what the offense and defense is doing is an entirly different thing...

I'm rambling. Sorry. :)

daedalus

05-31-2003, 03:33 AM

One thing I've always wanted to see was informational inaccuracies. I mean . . . the game that my scout saw you played, you might have had Wheaties for breakfast instead of Pop Tart and you were damn good or you might have a REALLY good night before and sleepwalked through the game. But the more that scout saw the player, the more accurate he/she would or should be. Within the scout's capability, of course.

Thinking in game term that might mean that a player with slightly above average hand (70) might've gotten either a grade of anywhere from average to very good (50 to 90, ratings plus or minus a random of 20) from an "average" scout (scouting ability of 80, leading to the 20). Another two games and that scout has managed to get a better read on that player and determine that the player has slightly above above average hands (55 to 85, ratings plus or minus a random of 15). And so on. Possibly with diminishing returns with more games. A better scout might not have been able to narrow down the range more, do so more quickly or not had such a wide range from the beginning.

And, of course, looking at the results will tell you something - that wide receiver you drafted in the 2st round because your scout thought he had above average hands ("72", according to your "average" scout) has been dropping 4-5 passes a game and probably doesn't really have above average hands (55 + 17 from the scout's misjudgement). Oops.

A possible solution to handling a new coach/manager would be to have a central scouting report. I don't know if the NFL has one but the NBA has a kind of a scouting organization run by Marty Blake (generally used for draft prospects). A new coach could be given a report based on either a scout with "average" scout ability (75 or 80) or a scout with the scouting ability that is the average of all employed scouts at the current time (who's going to listen to an unemployed scout sitting at the EDD office?). Once the coach is able to send out scouts, then that could would have his or her own set of report to work from.

Which, of course, leads to a whole new set of dilemma . . . do you allow a team to have multiple scouts (a la CM) or one single scouting group (a la FOF)? Because if you permit multiple scouts like CM, who's opinion do you use? One possibility would be to do an average of the grades given, although possibly with a modifier for more games seen (ie, if the "average" scout has seen 5 games but the "good" scout has seen only one). And, of course, also a possible modifier in case you've designated one a Lead Scout.

Oh. And, if at all possible, maybe a way to put configurations outside the game, particularly a text file (see: Sid Meier's or Brian Reynolds' games). I heart that.

Marc Vaughan

05-31-2003, 05:29 AM

sabotai - Regarding multi-player, I disagree with what you indicated about it being something you have to change the gameplay for.

CM has always been developed as a simulated world first (ie. a 'zero' player game) and then human interaction added into it via. screen hooks so that a human can do what the computer AI can.

In a similar fashion I'd always recommend building in multi-player gaming in the same manner, make any network code transparent to your game (this is especially easy if you are implementing a custom GUI system, for instance in CM3 you do the same thing to draw a screen ignoring whether the player is on a networked PC or a local host).

Obviously though your call and a good solo-player game is much better than a crap multi-player one ....

PS. If you want to discuss approaches to game dev feel free to drop me a mail :D

Eilim

05-31-2003, 06:14 AM

"Not to over-complicate things, which I tend to do , but any plans on the counter-play idea. You create a play that you use and eventually a counter play gets created by another team for it (like a defense to stop it). Sorry, now you can see how I led myself down that endless path when I started doing the play designing."

One way to give an illusion of sorts of "counter-plays" or a "learning AI" is to create a history of sorts of playcalls and give bonuses and negatives based on tendencies.

In a simple football engine I created a while back with a fixed amount of plays I kept a history of each play called, and the current situation. In the game engine itself there was a check for down and yardage, which it would then check against the "history" for how many times the current called play had been used in the game in the same situation, or if it had been called X times in a row. Based on this, certain stats were boosted to represent the players recognizing the play and adjusting.

Was a very simple engine and more a experiment in programming, but that was one of the features in it I was sproud of. Towards the end when I had added in a more advanced playcalling system I had even adjusted the history/AI checks to a complexity of checking current situation, formation, and players on the field. The big thing is playing with/tweaking the numbers to get a realistic representation.

On the singleplayer/multiplayer issue I understand your decision. I'm pretty much in the same boat with my own game projects but its more a design/programming experience issue than a core design issue for me. I'm just not skilled enough to make a game thats good both single and multiplayer yet.

-Eilim

oykib

05-31-2003, 07:02 AM

Play calling is essentially useless without a graphical representation. If you are not going to have graphics, then people will not really understand what's happening with plays.

It would probably be better to allow players come up with a detailed philosohpy. Preferably, you could create a gameplan wizard that could translate macro philosophy for people who don't wanna sit there for four hours like we had to in FBPro days.

A big thing would be owner personalities. That's one of the really big things in CM that it missing from the other sims.

stkelly52

05-31-2003, 08:59 AM

One thing that I would like to see on player ratings is if you were able to see their ratings, but the ratings are based only on their performance. Perhaps there could be 3 ways of looking at their ratings. You could see ratings based on the last season's performance, ratings based on this season's performance in practice (and training camp), ans ratings based on this season's game performance. This way you would not have an exact picture of their rating, but it would be close.

Taur

05-31-2003, 10:44 AM

Finaces, Finances, Finances. The "Front Office" Kind. Let me build, buy, and expand my empire. Or, face the dreaded bank when my loan is called.

Here is a cheap/gimiky idea:

I want a Salary. I may enter the league as an unknown coach with a 3yr-6mil deal, but after some success I will want a 3yr -10mil deal with incentives. After my second Superbowl win it will be a 5yr-25mil deal with even more incentives AND a GM package that gives me a percentage of the profits.

Many years later....
It is off to the Bank to finance a deal to purchase a team. Now as owner/GM/Head Coach you better watch the bottom line or your banker will give you a call.

Like the real world you can't actually spend more than you have.(loans and bonds)

BTW---Levels, levels, levels. Settings going from easiest to impossible. I like a challenge.

Airhog

05-31-2003, 11:37 AM

Look at how NCAA college football does recruiting. Even though it only gives you some stats like 40 time and bench press, its very easy to change this over and get a fairly good idea of how good a player is.

sabotai

05-31-2003, 11:50 AM

daedalus, I was thinking the same thing. You only see what your scouts see. And your scouts base their judgement on only what they see. So if you send your scout out to view (either at the game or on tape) a game and it's the only game any of your scouts see a certain WR, and he sucks that day, they won't think highly of him.

This of course will be offset by the media. If the buzz is that he's a great WR, you'll have to ask youself if your scouts saw in him things that will make him fail, or did they catch him on an off day? Or are his stats and skills over-inflated due to facing poor defensive teams all season? Or maybe he was being thrown passes by the back up QB that day, and his timing was off?

Marc, I would agree with you on multi-player if we were just talking network multi-player. What I was talking about was games like Fast Break Basketball and OOTP where players are doing their turns and sending in their files to the commish he has to handle all of the files and process them and send them back out. Of course, FTP support helps a lot, but the job of the commish in these leagues still takes time. And to make it worth their time, you have to make it easy on them and that forces you to make certain design decisions. Otherwise no one will create internet leagues using your game, and that'll be sales lost.

On the other side, if your game is not a competative single-player experience, you lose sales in that market. So in my opinion, you should just pick either single-player or multi-player and go with it.

Network (LAN) type multi-player allows you to stick with single-player design and development. I don't think design for internet support ala OOTP lets you do that. At least, I haven't seen it yet.

Eilim, I was thinking of adding something like that. Most likely all NFL teams do keep a record of what their opponents do in certain situations. So I want my sim to do that same (keep a record of play calling). How i do it is still up in the air. I could just add it to the play-by-play record or keep it in a different file.

oykib, I agree with plays their has to be some representation of the play. Which is why the simming part of the game is what I will work on first. If I can't get a satisfactory display of what is happening, I will fall back on an FOF and the previous FLS style game plans.

stkelly, everything you see as far as ratings will be through your scouts, so they will only be based on their observations. I could add in things like "He didn't do good last year, but he's been outstanding so far in training camp." Things like that (but obviously much more elaborate).

As far as numerical representation of attributes, this will probably be an option and show just several overal ratings. I dunno how I will do that yet.

Taur, it seems like what you are describing is more of a role-playing aspect instead of a simming game. :) Maybe in the future, if I ever do this, I'll add more RPG elements or a mode like in OOTP.

As far as difficulty settings, I don't like them. And in the end, everyone just ends up playing on the highest difficulty setting anyway. So my solution is not to create artificial challenges. Just make the best AI you can.

oykib

05-31-2003, 12:10 PM

I think that physical attributes should be clear and solid. Speed, strength, endurance... Those are things that a coach does not need to rely on a scout for. You should have a clear idea of where these guys rank. You can keep mental atrributes and certain skills hidden. But it'd just be annoiying if my scout misrepresented a player's speed to me.

You've already mentioned that you'd include 40 times and the like. So you've already thought of this. But that's just something that would annoy me about a game. So I wanted to clearly point it out.

sabotai

05-31-2003, 12:15 PM

Right, and the physical attributes will be clearly represented in the work out stats. And there will be plenty of those, so there will be virtually 0 ambiguity in those. When a guy is fast, you'll know he's fast. If he's strong, you'll see it in his reps. If a guy is agile, you'll see it in other work outs. Acceleration, vertical jumping, etc. will all be clearly represented in the work out stats.

It's just my opinion that having 40 times and shuffle times is more immersive than "Speed 84. Agility 79. Acceleration 90."

QuikSand

05-31-2003, 03:12 PM

I applaud your aggressive approach. You're striving to have each of these things done the "right" way, which is good. I hope you still have a workable project whwen it's all put together.

Karim

05-31-2003, 03:17 PM

A few thoughts...

I prefer attributes to scouting reports. That being said, only core "visible" attributes should be listed. We know if a guy is fast or slow. We don't know if a guy is consistent and this shouldn't be visible. We'd have to determine this through his actual play and scouting. There are all sorts of these dichotomies: A receiver can run routes but is he a big-game performer? This is the type of distinction I would like to see when deciding what to show and what to keep permanently hidden.

A realistic model of the salary cap and free agency. We all know that in FOF, there are no transition players and that franchise players and RFA's do not work in FOF the way they do in real life. THere are also no contract incentives and bonuses. What I want to do is have a true to life representation of the NFL free agency, salary cap and contracts.

That sounds absolutely awesome and would definitely be a hook for me. I'm much more interested in the general manager aspect of sports sims and would much rather prefer hiring the right coach for the team.

Sounds good...

vtbub

05-31-2003, 03:30 PM

Ability to import historical seasons would be my biggie.

Ability to call plays too, as long as the defense AI is tough and can learn.

Ability to actually play a game over multiplayer instead of endless exports/imports.

Whatever, Keep it up. I actually had your Tennis sim and thought that ws very good.

sabotai

05-31-2003, 05:13 PM

My tennis sim...damn that was a long time ago. And another one of my glorious failures. :)

mckerney

05-31-2003, 05:20 PM

No static positions, see Terrell Suggs or Corey Chavous. Just because a guy was one position part of his career should not mean that he stays at that postion his entire career. And don't make the mistake that was made in FOF in declining ratings for switching positions and losing ratings for switching back.

Is it really that uncommon to try guys at different positions in practice?

sabotai

05-31-2003, 05:26 PM

mckerney, it's probably uncommon in the pros, but probably not in college and it's definatly common in high school.

No, players will not increase or decrease their attributes with position switching. But they will not trained as well so it'll take awhile before the player is playing at top level in his new position. IOW, he needs time to learn it. That'll be determined by how smart he is, etc. etc.

mckerney

05-31-2003, 05:28 PM

Originally posted by stkelly52
One thing that I would like to see on player ratings is if you were able to see their ratings, but the ratings are based only on their performance. Perhaps there could be 3 ways of looking at their ratings. You could see ratings based on the last season's performance, ratings based on this season's performance in practice (and training camp), ans ratings based on this season's game performance. This way you would not have an exact picture of their rating, but it would be close.

What about a scouts opinion? Many things other than performance goes into judging players. If this years draft were to be run by game with just examing how player performed would Kyle Boller have gone in the first round? No, simply looking at statistics there would be several QBs selected before him. While Kliff Kingsbury may have had a better college performance numbers than Boller, scouts still thought Boller would be better. There are many cases of players being drafted on potential who didn't do all that much in college where guys who had great performance were left for the late rounds if drafted at all.

What needs to be done is find a way to factor in what scouts may see in players, though in a text only world with all the intangables other than performance could be quite difficult.

mckerney

05-31-2003, 05:38 PM

Originally posted by sabotai
mckerney, it's probably uncommon in the pros, but probably not in college and it's definatly common in high school.

No, players will not increase or decrease their attributes with position switching. But they will not trained as well so it'll take awhile before the player is playing at top level in his new position. IOW, he needs time to learn it. That'll be determined by how smart he is, etc. etc.

Major position changes may not occur, though I wouldn't say changing position changes are uncommon. Players switch defensive and offensive line spots, linebacker positions, between safety and corner, RB and FB, wide reciver and TE. Major changes such as WR to CB are uncommon, though they still do happen.

Also, as far as position designation goes it can change based on what system is being run. In FOF4 a guy may be designated RDE, though playing right end requires a differnt type of player with different responsabilities between the 3-4 and the 4-3.

As far as learning new positions, perhaps it's best to have a hattrick like experience ranking that is affected by the type of coach. Also, learning something for a position is not exclusive for that specific postions (i.e, a guy who learns to play rush end also gains some experience for playing a blitzing OLB in a 3-4, though obviously not as much as he learns for end, and his abilities still have an affect).

stkelly52

05-31-2003, 05:56 PM

Originally posted by mckerney
What about a scouts opinion? Many things other than performance goes into judging players. If this years draft were to be run by game with just examing how player performed would Kyle Boller have gone in the first round? No, simply looking at statistics there would be several QBs selected before him. While Kliff Kingsbury may have had a better college performance numbers than Boller, scouts still thought Boller would be better. There are many cases of players being drafted on potential who didn't do all that much in college where guys who had great performance were left for the late rounds if drafted at all.

What needs to be done is find a way to factor in what scouts may see in players, though in a text only world with all the intangables other than performance could be quite difficult.

I think that scouts should be used to help you determine the future of a player. So basically, I would like a report that says this is what the guy did last year, he is going to keep getting better.

Fido

05-31-2003, 06:39 PM

First off a question for you - are you planning this to be a solo effort or are you looking for help? If you want help, let me know, it sounds like you have a good plan for this.

On the attributes issue. I don't think you need to display much of anything. NFL coaches don't know that their top receiver has a 90 in hands and a 75 in pass routes. They know his 40 time, and that he caught 108 passes for 1200 yards last year. As for initial drafts, this can be done by simply creating back statistics for players (which I think should be done anyway. It bugs the crap out of me when you start a new game and see a player has been in the league for 10 years but has no stats. It's not that hard to determine what those stats would/may have been. This can also give an indication of players on their way up and players on their way down.

Another way would be to steal an idea from TDCBB and have scouting reports on every player. "This guy is simply a practive body", "This guy looks to be a starter right now". These would all be based off of your coaching staff's scouting ratings.

This is a brainstorming thread, so let me throw out some ideas of what I think would be cool to see.

College prestige. It seems like you are making a pro game, but college is a big part of that. Sure, Dartmouth College may have a star player every decade or so, but the NFL know that schools like Miami are where to go for talent. That should be reflected in the game.

Play familiarity. Something that seems lacking in simulatio games is that you can add plays to your playbook and have the team know them cold immediately. I think practicing plays is more realistic, and the team's effectiveness when running that play should be based on how many times they've run it oin the past.

Blocking/coverage schemes. If you know that your opponent's Defensive end is a pass rushing monster, you should be able to easily set it to always double team him on pass protection schemes. Same with coverage. You may want to always double cover their star receiver.

Emotion. Players returning to play the team that refused to sign them the previous year will generally be emotional going into the game. Some people step up, some crumble. But it should be a factor.

Keep play calling out of it. I know a lot of people like it, but it gives the human an unfair advantage. You'll need a really strong AI to match wits with a human player, and since this is single player only, there's no alternative once a person knows how to beat the computer coaches.

Open file format. Can't stress this one enough. You are not going to be able to include everything that every user will want into the game, you don't have the resources for that. Doccumenting your data file format and making that document available will allow users to easily write third party utilities to fill the voids. There is so much that I would like to do in the next version of TCY helper that is just being a bear because Jim has decided not to provide any assistance in the format of the files. People will hex edit the files and figure it out, keeping it secret will not prevent that (unless you want to encrypt them, but why would you want to do that)?

K.I.S.S.. (Keep It Simple, Stupid) More of a development suggestion. Plan for the features that you want, but keep the game simple. Work on the most important part of it first - the simulation engine. Then build from there. A beautiful game with lots of bells and whistles will not be used if the simulation engine is crap. Compartamentalize everything and have well though-out interfaces into each so you can just plug stuff in as you add it to the application.

It looks like you have a good plan in place already, and are open to comments, which is always good to see. I look forward to seeing where this project leads.

tucker342

05-31-2003, 07:32 PM

good advice Fido.

sabotai

05-31-2003, 07:50 PM

mckerney, I see what you're saying. Yeah, minor position changes, and changes along the o-line are common. Big ones are not though. And I agree with the new position learning.

Fido,

Actually, I don't have any kind of plan right now. Just some general ideas. I only have a couple of pages scribbled with ideas. And at this point, it's a solo project simply because I don't know when I'll actually write out a design doc, no idea when I'll start, no idea how I'll end up doing most of it, etc. But if I ever want to branch out and get some help with this, this forum will be the first to know. (And this will never be an open source project)

As for you ideas, it's like you're reading my mind. Except for play calling. :)

Blocking/Coverage scheme is a must. Play familiarity is a must. College prestige is a definate must. Open File format will happen. I've always been a fan of 3rd party developers. I've always thought they helped games be better.

The sim engine will definatly be where I start. Everythign else is really a matter of good AI design, good programming. The sim engine, however, requires more. And since the AI outside of the sim engine relies on how the sim engine is done (game plan decisions being the biggest). I hope to create a physics based sim engine. But I don't know if I can do it. I also have to have AI that is based on real football.

If I can't do it, then I fall back on attribute based engine with an FOF style game plan.

Since so much relies on which way I do it, the sim engine will be the first part done.

As for play calling, as long as I put reasonable restrictions on it and play design, I think it'll be fine. Unlike FBPro, where you could draw anything, this will be restricted to predetermined routes, running directions, blocking schemes, etc. You won't be able to draw in your own stuff to beat the AI. And as far as play calling, as long as I make players stick with there game plans, it shouldn't be a problem.

Like say you have 6 plays in your game plan to deal with goalline situations. You'll only be allowed to call those 6 plays.

Also, I think it wouldn't be too much of a problem to log a players play call tendancies and have the AI 'consult' it to try to predict what the player will do.

The reason, I think, that in FOF you eventually "break" it is because 1) I don't think there's much to the AI decision making on D 2) The player can call any of a TON of plays. It just doesn't work like that in real life. Coaches stick to their game plans because that's what they practiced and that's what the players know.

So in short, as long as I limit certain things, and as long as I create a good play calling AI, I don't think it'll be much of a problem. And if it is, I can always take it out. :)

sabotai

05-31-2003, 07:55 PM

dola,

And this will stay in the brainstorming/loose design stage for awhile. Many of you may remember I was programming FLS in Delphi. This will be C++. Yes, I'm a convert. :)

My next project, which could be 1 of 3 different ideas I have, will be my next big thing (not really BIG, they are just simple ideas). This I'll use to better familaize myself with certain aspects of C++ programming. Although I now know the language very well, I haven't really done anything in it (except write a bunch of tutorials on it. www.bearklaw.com shameless plug).

Anyway, since I haven't really done anything in C++ yet, I plan on doing 1 or 2 small-moderate projects first to 'practice' my skillz.

kparker15

06-01-2003, 12:41 AM

This is a great thread. I am very familiar with football and have been playing football games since the original fbpro. I have played fbpro, FOF, Action PC football, madden, ect.

All of these titles have something very good. The best model for a pure coaching simulation is Action PC Football. Fbpro can be beat (even with the VPNFL files -- those who know this game will understand what I am talking about). After playing TCY, I wondered: What makes a good football game true to life. Here is what I think.

I enjoy TCY in that you really do not have an active role in the play calling itself. It is somewhat suspensful to see if your recriuting efforts pay off (especially with the new viewer). No matter what you think, if a human has control in the playcalling, the computer can be beaten. Even in Action PC football this is the case (most of the time).

SO this brings us to the question: What goes in to gameplanning and preparing for a football game?

First, you scout your opponent. The game of football is like a chess match. Teams have players who are strong at certain positions and are weak at others. Good coaches will take advantage of these mismatches and utilize plays to exploit them.

The first place this comes into play is on the line. Let's say you are planning your offense for the upcoming game. The game would give you a scouting report of the defensive linemen. It may read
"DE Kearse has excellent speed and tends to rush to the outside; THe right side of the defensive line seems to be weaker."

"Tennessee's LB are quick to react to the ball, but are often fooled by play-action"

"The right CB is slow and has been beaten deep several times this year. The left CB usually is matched with the opponents best reciever."

"The overall defensive scheme for the Titans is a 4-3. The Titans have blitzed 45% of the time on third down this year. Tennesse is giving up an average of 17 ppg and is ranked 3rd in rushing and 17th in pass defense."

I have yet to see a game with this kind of in depth info on a team. The stats should be dynamic and change as the year progresses.

Given the defensive scouting information, you, the player, can choose what type of plays can fill your gameplay. At the beginning of the season, you have to declare your primary offense (run and shoot, west coast, balanced, ect.). Each offense will have variations which can be manipulated as the year progresses. Certain plays will feature certain players and certain plays will work better against zone or man coverages. It is up to you to pick the plays to fill your plan.

So, we have covered day one of the game planning process. Scouting and finding weaknesses and then determining which plays you want to use in your plan. Next, the teams will practice the plays during the week and review film. To this end, you can see what kind of defensive or offensive plays are in the opponents plan (watching film). Go into the simulator (ala fbpro practice mode, but more simple) and try a few offensive plays against a scout defense to see the result (practice). The injured players will rest, get rehab, ect.

Now, you have your plan set (25 runs, 25 passes; keep it simple, nobody wants to go into a game with more than 50-60 plays. I know NFL playbooks are 500 pages long, but the games total number of plays would fill this book. However, your team will only have 50-60 plays for each half.

Next, it is gameday. I agree with Fido, playcalling should be left up to the computer or better yet the coach you hire to run your team. I know many of those on this board will disagree, but hear me out.

You the gamer, have prepared for the game by scouting and maximizing your matchups. You have put in plays that feature certain players and take advantage of the defense. Your coach (or offensive coordinator) will call the plays. No matter how advanced computers are, if a human brain is involved in calling the plays, the human will have the advantage. I think we all have seen this in all games we play.

This is the fun part. You watch the game unfold in a nice 2-D interface that allows you to sub players and make adjustments to your plan at halftime based on what the defense is doing. The computer AI also is dynamic in that, it may switch from the norm and come out in predominantly man (rather than zone). Or run a QB sweep the first three plays from scrimage ( as Steve Young did in the 1994 NFC championship versus Dallas). As you watch the game, this is the essence of football. You spend all this time preparing and now it is up to your team to put that preparation onto the field of play. It is here where you see if the friuts of your labor pays off.

Yes, a few years ago, I too would have been adamant about playcalling. This is indeed a big part of the game. But in the game you match your wits against a human opponent. Being an old man with two young children, I do not have time for online play. In solo play, the only way to make the experience a realistic and enjoyable on (IMHO) is to take the individual playcalling out of the equation.

Just some thought, I could go on, but it is late. Good luck on the game, it sounds like it will be a well thought, and productive process.

Ken Parker

Celeval

06-01-2003, 01:44 AM

Originally posted by Fido
NFL coaches don't know that their top receiver has a 90 in hands and a 75 in pass routes. They know his 40 time, and that he caught 108 passes for 1200 yards last year.

Well, no, but they do know, for instance, that Keyshawn Johnson has better hands than, say, Warren Sapp. And that Jerry Rice has a better work ethic than Randy Moss. And Peter Warrick runs better routes than Akili Smith. The question becomes what ratings represent - in real life, coaches find this out from watching a player, from their scouts, from their practice performance, from all kinds of things. The ratings are we what we've used to represent this, because computer simulations are by default number-based games. If not ratings, there should be some method of finding things out about players other than stats - I don't need to put Jeff Faine in at quarterback for three games (having him go 3-18, 32 yards, 5 interceptions) to learn that he shouldn't be one.

Kevin

sabotai

06-01-2003, 02:03 AM

If you don't want to play call, you can leave it up to the coach. It'll be there for those who want it. Everyone's happy?

I understand the play calling arguments, and for th emost part I agree with them.

There are several reasons I can think off of the top of my head as to why player's "break" the AI in play calling.

1) FBPro style play design. You got to draw the routes and everything else. Very easy to create "AI Busting" plays. The fix: Don't allow that kind of flexabilty. Keep the plays within the standards of real play design. For passes, must pick from list of premade routes. Runs go to specific holes. Blockers block according to a premade scheme. No AI busting plays.

2) Sim AI. This goes to FBPro again, and most, if not all, graphics games. The AI was just broke. It did not perform according to the logic, assignments and routines of real football. The fix: Make the AI respond and act exactly the way they do in real life.

3) Play calling. In real life, coaches have game plans. Each situation has a list of several plays under them. Coaches stick to these plays (rarely, maybe 1 play every game or two they take one from their "trick" play list). In every single game out there, you get to pick any play you want. This allows you to call plays on 2nd and 2 that you never would in real life. The AI is thus unprepared. The fix: 2 parts. Make sure the AI is aware of the possiblity that the team might go long on 2nd and 2, especially if they have in the past. 2) Make players stick to their game plans. No calling 3rd and long plays on 1st and ten (unless it is listed under both).

4) Unprepared AI. In most games I have seen, the AI simply does not look into the play calling tendancies of its opponents. They stick to premade path. If you throw a pass 87% of the time on 2nd and 5, then the AI needs to be aware o fthat and prepare themselves for a pass. In most games, the AI just picks a base defense, preparing for both. This allows you to throw all day long on 2nd and medium and the AI never adjusts nor does your next opponent prepare. Fix: Keep track of play calling tendancies of the human (and cpu) teams and make said info available. Have play calling based on it.

5) Unprepared AI part 2: The AI , in most games I have seen, never seem to prepare correctly. They stick to their premade paths. They don't try to exploit weaknesses and they don't try to stay away from strengths. If a team has the best run defense, but piss poor defensive backs, then the AI should realize this and throw it all day. Then after they establish the pass, then go to the run to keep them honest. I have never seen an AI do this. The fix: Make sure that in preparing for the game, the AI analyses their opponents for weaknesses and try to exploit them

6) Play familairty. In every game out there, you stick a play in your playbook, and your team knows it 100%. It doesn't work like this. You must practice your plays before you know them. And if you don't practice a play for a long time, you forget it. No sports sim takes this into account. A coach has a certain amount of plays he runs. He teaches them to his players. He doesn't just come up with 40 new plays and throw them into his game plan on game day. But games allow you to do this. The fix: Make sure teams practice plays before they can even use them, and make them practice it a few times before they 100% get it.

7) AI play picking. No game out there that has playbooks creates an AI that picks their own plays for the game plan. They stick to a game plan that is preprogrammed, and it never changes. This allows the player to easily reconize offensive and defensive plays. Whenever I'm playing Madden, and I see something for the second or third time, I catch on, and I nail it. What would be cool is if the AI puts in a play that does the same thing, but the play goes the opposite direction. There I am, putting my line and linebackers to one side, and bam, the play goes the other for a big gain. But this never happens. The Fix: Allow the AI to pick plays to put in a game plan, and to reconize plays that may not work anymore. Also, allow it to create counter plays to try to fool teams.

It's 3 am. I know of some other things I have thought of, but I have forgotten them right now. :)

The point being. I do not think it is impossible to create an AI that is and always will be competative with a human player. It'll take a lot of work and a lot of time, but it can be done. If they can create an AI to win at chessat a championship level !! :) Football at the casual level really shouldn't be impossible.

Ambitious? Yes. Over my head? Probably. If I fail can I easily just fall back to a position of no play calling? You bet your ass I can. :)

sabotai

06-01-2003, 02:10 AM

Celeval, have no fear. I wouldn't just throw people in without giving them anything. They'll be detailed scouting reports, work out stats, history, etc. that will be given to you.

"Jerry Rice has a wonderful work ethic and attitude that makes players around him better." vs. "Randy Moss has a poor attitude that might make him a liabilty come game time. He has a history of giving up on plays that don't feature him and he does not give convincing run blocks."

Of course, it'll be much more detailed than that going into every area. I will do my best to give all the information that is needed by the player in a narrative way instead of a numerical way.

Fido

06-01-2003, 07:39 AM

Originally posted by Celeval
Well, no, but they do know, for instance, that Keyshawn Johnson has better hands than, say, Warren Sapp.And that Jerry Rice has a better work ethic than Randy Moss. And Peter Warrick runs better routes than Akili Smith.

Ok, but you'll only know that Keyshard has good hands ad that Sapp has poor hands. That's all you need. If you compare Moss and Rice, all you know is that they both have great hands. There's no reason to know taht one is arted 98 while the other is 95. Its providing more accuracy than is available in reality.

The examples you give are all based off of extreme examples. I am not saying that you should not know the difference between Kevin Faulk and Marshall Faulk. But you should not know exactly how much better one is. Only that one is great and that one is average, and their season stats and practice perfromance should demonstrate this.

Originally posted by Celeval
The question becomes what ratings represent - in real life, coaches find this out from watching a player, from their scouts, from their practice performance, from all kinds of things. The ratings are we what we've used to represent this, because computer simulations are by default number-based games. If not ratings, there should be some method of finding things out about players other than stats - I don't need to put Jeff Faine in at quarterback for three games (having him go 3-18, 32 yards, 5 interceptions) to learn that he shouldn't be one.

Right, that is where you call a meeting of yoru caoches and ask your quarterback what he thinks of your staff. That computer person will have acces to the real behind the scenes attributes, but will skew them based on his/her personal preferences (a QB coach who prefers drop back passers might prefer Manning over Vick) and scouting ability.

I'm not saying the user should have to guess, only that the information should not be handed to you.

Fido

06-01-2003, 07:56 AM

Sabotai -

I think that your system will work, but it will have varrying levels of effectiveness. The AI will get better and better as the career progresses as the computer develops a larger set of play caling history. At the start of a career, the computer will just be guessing, by your fifth season, it should know your tendencies cold.

One other thing you shoudl consider is the contents of the player's playbook (as demonstrated by previous games). If the player has a single play in the playbook that is run out of a given formation, then the computer should e ready to stop it. If your only play out of the I formation is a 46 power, then the computer should be ready to stop it. If you wait until the superbowl to bring out your play action pass out of that formation it should have huge success.

One thing that you might want to consider would be to make more job options available to the player. If someone REALLY wants to call plays, let them be an offensive coordinator which will let them do that (on offense) but will limit what else they can do (can only make suggestions on draft choices and depth charts). This could also make for an interresting multiplayer option (if you ever decide to do that) - a group of four or so players could each take a position on a given team.

sabotai

06-01-2003, 12:11 PM

"If your only play out of the I formation is a 46 power, then the computer should be ready to stop it. If you wait until the superbowl to bring out your play action pass out of that formation it should have huge success."

Exactly. And I want to AI to be able to do the same to you. Which is one of the main things AI's don't do in sports games. Surprise you, because they always follow a specific pattern.

"One thing that you might want to consider would be to make more job options available to the player. If someone REALLY wants to call plays, let them be an offensive coordinator which will let them do that (on offense) but will limit what else they can do (can only make suggestions on draft choices and depth charts). This could also make for an interresting multiplayer option (if you ever decide to do that) - a group of four or so players could each take a position on a given team."

Good ideas, but I'm not sure if I want to go this far in the first version. Basically I'll just have a checklist of things you can hand off to the AI for your team. If you want the AI to handle play calling and other coaches stuff, then just check the box for the AI to handle this.

With jobs and stuff, I think those things which are more role-playing type features will get pushed back. Got to have a reason to make a sequal, right? :D

44Niners

06-01-2003, 05:22 PM

Here are a few thoughts:

First, and I can’t emphasize this enough – everywhere it is possible, make options a toggle. It ends all the debate of see the rating, don’t see the ratings, call the plays, don’t call the play, etc, etc. The more customizable it is the more people enjoy it (especially since it is single player) and the more successful it will be. I have a garage full of sports games that I really liked, but there were one or two things I would have liked to be able to change.

Second and along the same lines, make the players, logo, look and feel 100% customizable ala OOTP. To me the greatest part of OOTP is if I am in the 3rd inning of the game and I think the opposing RF’s arm rating isn’t high enough, I can change it right then and there. Does that mean I can change all of my teams rating and win 100-0, sure it does. I am not sure why anyone would want to, but if they did who cares, it is their game – let them change whatever they want

Ratings –The one thing I never see represented in sports sims is a rating for heart/desire/attitude. At least in football these things make a much bigger difference then any other rating. Not to say a guy that runs a 5.2 40 can be a NFL receiver if he has a lot of heart. But the difference between a player such as say Jerry Rice and Randy Moss is their attitude. The way they prepare on and off season. Practice habits, attention to detail, etc,etc. Essentially the vast majority of players in the NFL are big, fast and strong. It is their desire, effort and preparation that makes the difference.