FOF: Finding a "formula" for Game MVP

[QuikSand]

There has been some discussion about how the FOF4 game selects the MVP for a given game, most visibly the championship game.

Strikes me that this is a potentially ripe topic for some legitimate discussion. Maybe we, here, can come up with some forumlas that would provide some generally sensible outcomes in this regard. Who knows, maybe we could even get these ideas into a future FOF4 patch.

So, let's have at it... how do you decide who gets game MVP, just based on game-end stats?

[QuikSand]

Okay, a dozen looks and no responses... maybe this needs a shove.

I'll start off with a pretty simple start - a "points" system for a RB. I am arbitrarily going to come up with a system that awards points in pretty small increments... so that a player with a truly excellent game wiould end up with something like 10 MVP points.

Here's my first cut:

RB points = the sum of:
(each item rounded down to an integer)
1 point for each 25 total yards gained over 50
1 point for each 5 carries over 20
1 point for each 2 catchews over 2
2 points for each TD scored
-1 for each fumble lost to other team

So, if you RB has this day:

22 carries for 94 yards, 1 TD
5 catches for 41 yards, 1 TD

He gets 3+1+4 = 8 MVP points, which ought to make him a pretty serious contender for the game MVP honors.


Now, without necessarily accepting my particular formula (we can debate how much to weight various components) - but accepting my general scoring concept (that 10 points represents a "great game") - what else can we come up with?

[Dutch]

You are asking for a lot to come up with a true Game MVP based on a formula that could be entered into the game.

Take for instance the latest Super Bowl. Dexter Jackson received MVP honors for making two key interceptions (where he visibly attacked the football unlike a couple of years ago when the Steeler DB picked off a couple of arrant passes that just seemed to fall to him). Dwight Smith intercepted two passes and ran them back for TD's but both were at relatively inopportune times as far as being "valuable" int's and td's.

How about a team that is down by 21 and starts to air it out in the final quarter? The QB may have only thrown for 100 yards in the first 3 quarters but gains another 200 in the final quarter and still loses 21-0. He should not even be considered for the MVP.

Obviously this would be hard to add logic to.

So perhaps we could think of easier ways for it to be programmed.

Some ideas would include giving all winning players a 50% bonus to their scores. That way if a losing player does put up great stats, he really earns that award because he outperforms even the best player's stats on the winning team by a factor of 1 and a half.

Also, if any player scores a touchdown while the score is within a 7 point range, that player gets a 25% bonus to his individual score.

But that's already making is more difficult again with logic that could easily become flawed.

[QuikSand]

I think you're right, Dutch... I have no aspirations that a forumla could account for every real-world circumstance and bit of situational logic. Asking that much is probably beyond this game. That's why I am aiming more reaosnably... shooting for "forumlas that would provide some generally sensible outcomes."

It's not going to be perfect (and how one could hope to draw instruction from this particular SuperBowl is beyond me... I'm still not sure for whom I would have voted, probably Simeon Rice). But a fairly simple forumla (maybe, hint hint, one that is not dependent on player position) might be reasonable. That may, in fact, be how the game already works - it's just that the formula needs tweaking.

Your idea about a % bonus based on being from the winning team... I like that. I had imagined a flat point or two (on this system), but a percentage (whatever seems right) might be even better.

[wade moore]

Quick,

I'm not much able to develop formulas, so I'll suggest general suggestions to the basis of your formula(s)..

Perhaps for certain stats (TDs in particular) maybe it is important to figure in what % of their teams total production a person had, rather than the actual #'s? Or a combination of the two? For instance.. in a 10-3 game, 1 TD is very important and could mean the mvp..

[cthomer5000]

I think *the* winning score in a game decided by less than 14 points should also be worth at least double.

[cthomer5000]

things that help a player's MVP cause:

• scoring the game winning score
• being on the winning team
• scores that occur while it is a close game (the definition of close game can be debated)

I believe all of those should be positive modifiers applied to each score/stat that they record.

should a touchdown be of universal value? Maybe a defensive touchdown should be worth more than an offensive one? They're more infrequent, and arguable have a bigger impact on the game. Going from defending against a score to scoring yourself on one play is a large momentum changer.

[wishbone]

I think the key to actually having this implemented is to use stats the game already records inthe game log or in the box score.

Rush of 20+ yds 1 point
Reception of 40+ yds 1 point
Completion of 40+ yds 1 point
Key run block .5 point

Fumble -2 points
Interception -2 points

TD Rush 2 points
TD Reception 2 points
TD Pass 2 points
FG 1 point

fumble recovery 2 points
force fumble 1 points
Tackle for loss .5 point
Int 2 points
sack 1 point
force fumble 1 points
Defensive TD 2 points
pass defensed .5 points

1 point per 100 yds passing, rushing or receiving
2 points if you break a record

I think the "winner bonus" should be dependant on the margin of victory.
0-2 points 25%
3-7 points 50%
8-14 points 60%
15-21 75%
22+ 100%

I don't know if a kicker or punter would ever be considered the game MVP so I don't know about even looking at them.

[wade moore]

bump

[Blackadar]

Originally posted by Dutch
You are asking for a lot to come up with a true Game MVP based on a formula that could be entered into the game.

Take for instance the latest Super Bowl. Dexter Jackson received MVP honors for making two key interceptions (where he visibly attacked the football unlike a couple of years ago when the Steeler DB picked off a couple of arrant passes that just seemed to fall to him).

That was Larry Brown for the Cowboys against the Steelers.

[JonInMiddleGA]

Originally posted by wishbone I don't know if a kicker or punter would ever be considered the game MVP so I don't know about even looking at them.

I think kickers have to be in there somewhere though. Off-hand example would be a K who goes 5-for-5 on FG's & hits a time expiring game-winner in a 15-14 victory.

I agree with you about punter's in the context of a sim (I've had a punter named PotG twice in my years annoucing HS football though)

[sabotai]

I'm not sure you would rule out kickers. (As Jon has pointed out).

I think Punters are out, definately. I also thin OL is out too. A center, guard or tackle will never be given MVP honors.

[Raven]

Originally posted by QuikSand

RB points = the sum of:
(each item rounded down to an integer)
1 point for each 25 total yards gained over 50
1 point for each 5 carries over 20
1 point for each 2 catchews over 2
2 points for each TD scored
-1 for each fumble lost to other team

So, if you RB has this day:

22 carries for 94 yards, 1 TD
5 catches for 41 yards, 1 TD

He gets 3+1+4 = 8 MVP points, which ought to make him a pretty serious contender for the game MVP honors.

[/b]

I think this formula highly overvalues receptions, and undervalues TDs.


I'd use something more similar to this.....

For RBs
30-50 rushing/receiving yards = 1 pt
every 10 yards thereafter = 0.6 pts
TD rushing = 3 pts
Every 8 rushing attempts = .3 pts
Every 3 receptions = .3 pts
Fumbles lost = -1 pts

SO...
22 carries (.6 pts) for 94 yards (3.4 pts), 1 TD (3)
5 catches (.3) for 41 yards (1 pt), 1 TD (3)

=11.3 points (instead of 8)

[sabotai]

The problem with TD values is that they quickly become over and under valued.

1 TD is not that impressive unless it is the only or one of the only scored. 2 is impressive unless it is one of 6 scored.

TD values should be more exponetially AND be relative to the total # of TDs scored.

Heres something (and it's off the top of my head so it's going to suck...)

Take the RB above.

22 carries. 94 yrds, 1 TD
5 catchs, 41 yards, 1 TD

Let's say you start with a 2 base point for TDs
The first TD is worth 2 points. Let's increase the value of TDs 50% each time. 50% of 2 is 1. So we increase it 1 point. That makes the second TD worth 3 points. THis way, we give a little bonus to those who make more TDs. It'll make it harder for other players who don't score as much to make up the difference in other areas, but doesn't over value that first TD to the point that the other stats don't matter.

And we keep doing this. 50% of 3 points is 1.5 so the 3rd TD scored would be 3 + 1.5 = 4.5 point. This puts a lot of points into that 3rd TD. And a player who scores 3 TDs in a game will most likely get the MVP honors, almost regardless of other stats.

And then we make it dependant on the other TDs scored.

Let's say 3 total TDs were scored in the game above. The RB cored two of them. he accounted for 66(.666666)% of the total TDs.

Now if we took 5 and multiplied it by 2/3rds, we'd get too little a value (3.3333).

So let's do something else. Why not -5% of the TD score for every TD scored in a game that the player did not score.

The RB above scored of the 3 TDs in the game. His TD score is 5 (2+3), the other player's score is just 2.

The RB above would have 5% taken off and the other player would have 10% taken off. 5 * 0.95 = 4.75 , 2 * 0.90 = 1.8

Hmm, now that I look at this, it may not be all that good. 5-2 = 3. 4.75 - 1.8 = 2.95. The difference is smaller.

SO let's drop the % side of it. Why not just make it something more concrete. We'll - 0.5 points for each TD the player did not score.

The RB above would have 5 - 0.5 = 4.5 And the other player would have 2 - 1.0 = 1.

Since the RB scored two of the three TDs in the game, it would make sence that his 2 Tds would greatly outweigh the 1 TD of the other player.

But what if the RB scored 2 of 7 tds scored? Then his score would be 5 - (7 * 0.5) = 5 - 3.5 = 1.5

We completely devalue the TDs the RB scored because his TD production was only 28% (2 / 7 roughly) instead of the 67% from above. THis will make the MVP honors go to the player with the biggest day in terms of other stats. Liek the RB above could make those points back if he had several 20+ yard runs. Since everyone has their TD total devalued (unless someone scored 5), this would give the RB a jump on the rest.

However, if he just had merely a productive day instead of an explosive day (94 / 22 = 4.27 yard per carry. If his longest run of the day was only 7 or 8 yards, the a WR who made a couple of big receptions or a QB who made long throws the entire day would stand out against him.

Productive players get the MVP in low-scoring/close games. It's the explosive players that get the MVP in the high-scoring games.

[Vince]

I like the idea of valuing TDs based upon the amount of TDs scored in a game.

[sabotai]

bump - of course, once I ring in on the subject, the thread dies.

[QuikSand]

Hmmm... I'm not sure I buy this idea that "MVP points" need to be adjusted based on the total points scored overall in the game.

In a simple system, you might have a game with a 10-6 final score. With relatively few big plays and other things that garner "MVP points" you might have the final analysis of MVP points look like this (where W=winning team and L=losing team):

RB-W = 5 pts
WR-L = 4 pts
QB-W = 3 pts
QB-L = 3 pts
etc.

In another game, where the final score was 41-28, with lots of TD being socred and lots of MVP point bering handed out, the final tally might be:

RB-W = 11 pts
WR-L = 9 pts
QB-W = 6 pts
QB-L = 5 pts
etc.


In the first case, 5 MVP points is the highest in the game, so that wins the day. In the other game, 5 points is only good enough for an also-ran position, since others had bigger efforts.

It seems like a self-adjusting system... simply award points based on performance, and whoever gets the most wins the award. I think this mechanism works pretty implicitly in the scoring system I'm suggesting - that many of the implied goals of sabotai's system above are going to be reached anyway. (The 2TD RB in a game with only 3 TD in the total will be MVP anyway, since there just aren't that many points to go around... the 2TD RB in a 7 TD game has a lesser shot since there will be more other point out there belonging to someone else)

[sabotai]

QS,

The whole idea is where the emphasis is. In a high scoring game, TDs should not be the emphasis of the MVP award, big plays should, because so many TDs are scored, people don't really remember them. What they remember are the big plays. In a low scoring game, the TDs scored should be the emphasis, because the few socres is what people remember. This will adjust the value of each of the other points relative to the TD value.

Take the example above. An RB who gets 2 TDs when his team scores 5 others will get a total of 1.5 which makes it 0.75 points per TD. Now let's say we give 1 point for every play 20 yards and more. This makes the value of the big plays higher than the TDs.

But in the game where the RB scored 2 of 3 TDs, he gives a total value of 4.5 for the 2 tds, making it 2.25 points per TD. This makes the TD value much higher than the big play point.

I think the whole point of this is the hand out the MVP award as realistically as we can. A static point system basically just gives it to the guy with the best stats. And the guy who gets the best stats doesn't always get the MVP.

[QuikSand]

Originally posted by sabotai
I think the whole point of this is the hand out the MVP award as realistically as we can. A static point system basically just gives it to the guy with the best stats. And the guy who gets the best stats doesn't always get the MVP.

I can't really disagree with the sentiment, but I think there's also a certain elegenace to doing this within the means that are readily available. My original thinking was that we use the boxscore, and the data contained therein - that's it. Of course, in real life, we'd look at plays that changed momentum, and other intangibles... "big tackles" and "timely catches" and so forth. We're simply not going to have that in a text sim output.

oductive players get the MVP in low-scoring/close games. It's the explosive players that get the MVP in the high-scoring games.

On a little further thought, I really don't buy your assertion that somehow "big plays" (whatever that means, exactly) become the defining point in a high scoring game. I don't think there's really a deliniation between the way we pick an MVP in high-scoring versus low-scoring games. Call it ptoductivity, call it "the best stats" - whatever you want... the MVP is the guy who made the most difference.

I think we get closer with a simple formula (understanding that this will miss certain "human elements" that we'd measure if we really watched the game in person) than we do by backing out calculations of different things due to the "type" of game involved.

[EagleFan]

I don't have FOF4 yet. Is it better at picking game MVP's than the previous versions. The formula seems a little out of whack at times, like in a 30 point blowout and a defensive player that had an interception gets the MVP over the QB that threw for 300+ yards and 3 TD's. These aren't exact numbers but the previous games have seemed to over-value some defensive stats at times.

[sabotai]

"I don't have FOF4 yet. Is it better at picking game MVP's than the previous versions."

No, in FOF4, a player on the losing team routinely wins the game MVP. I don't even pay attention to it anymore.

"Of course, in real life, we'd look at plays that changed momentum, and other intangibles... "big tackles" and "timely catches" and so forth. We're simply not going to have that in a text sim output."

Why wouldn't we? Instead of looking at box scores, we can game logs. Look at primelord's utility that he's working on in the dynasty reports. He's grabbing good stuff off the game logs.

I do think a simple point system like the one you suggested would be a good way of handling game MVPs. It would assure that good, productive players go the MVP. I just don't think that in real life, game MVPs always go to the productive players.

Look at the Super Bowl. Michael Pittman had 29 carries for 124 yards. But Dexter Jackson got the MVP, even though he only had 2 tackles. He got the MVP because of the 2 INTS he got. They were early, game altering plays that guided the game to the blowout it became.

But if we spin things around and make the Super Bowl a 21-20 win for TB, does Jackson still get the MVP? I don't think so. He did run one INT back for a TD. If he ran the second for a TD, I think he gets it. But in a close game, I don't think a defensive player with 2 INTs and 1 taken for a TD gets the MVP. Unless of course, the TD won the game. In fact, the player who scored the game winning TD, as long as he wasn't some backup HB who had 2 total carries, would get the MVP.

And this is why I don't think a simple point system gives us realistic MVP results. The MVP depends so much on situations, end score, the impact of specific plays, etc. Not just box scores.

I'm not saying a simple point system is a total failure. In fact, for a test sim, a point system will probably be "good enough". But if the goal is to develope a system to realistically pick players for game MVPs, I don't think a simple point system will work.

[sabotai]

After thinking about this, I think you're right QS, in a way. I don't think it's impossible to get a text sim to realistically pick MVPs, but to do so would require weeks of design, implementaton and testing. For a text sim, that's time best served elsewhere. As for test sims, simple points systems would be the best bet. They're fast to design and implement.

[wade moore]

I somehow missed some of the conversation here..

I agree with QS.. when I pointed out the TDs scored I didn't think it out thoroughly realizing that it would balance itself out.. I do believe we need to add the % for the winning team though... Perhaps we could add a bonus for the last person to score in a game where Final ScoreW-Final ScoreL <=7 ? Does that make sense?

Beyond that, I think I like QS's design.. perhaps we should put it to the test after he makes a final setup? Maybe quicksim a couple of weeks and see if the MVPs make sense for a week or two of games?

[albionmoonlight]

I like the set-up and the idea. I think that there should be extra points given for extraordinary performance. For example, instead of 1 point for every 25 yds. over 50, say 1 point for every 25 yds. over 50 and a bonus 1 point for every 25 yds. over 150, etc.

I think that someone with more time than me could devise a system by looking at a set of boxscores, deciding who the MVP should be, and making a formula that comes out right.

Maybe a first step would be to post a couple dozen box scores on the board and having people state who they think MVP should be based on the box scores. Then, once you have that good consensus, devise a formula that will produce those MVPs.

[SunDancer]

How is this coming along?

[albionmoonlight]

Hey, I had a perfectly good thread-kill going and you ruined it

[SunDancer]

Haha. Just something I had some questions in the game itself. Just wonder if you guys came up with good solutions

[QuikSand]

I think the conversation just died of its own weight. Might have been a Hattrick match that day or something.