The Birth of the Treefrogs (the first Desktop Dynasties dynasty)

[GoldenCrest Games]

Standings

You might notice the change in MVP scoring this season. I rebalanced the scoring system, based off last season's final stats. It should be a lot more fair to non-QBs and non-KR now.

Wow, 3 weeks in and Luna has already started dominating the passing leaderboard. That sure didn't take long! Also, we have an Urquhart sighting!

[jct32]

Luna leads the league in passing, #8 in MVP. C'mon man!

Anyways, looks like I have missed a couple of updates. Season is off to a nice start. It will be interesting to see how the kicker acquisition plays out in the long run.

Additionally that defense is looking decent this year!

[GoldenCrest Games]

He's gotta step it up on defense!

[jct32]

Originally Posted by GoldenCrest Games

He's gotta step it up on defense!

300 yards more than anyone else and they want him to play defense? This ain't Neon Deion!

[GoldenCrest Games]

Most of my games seem to go about like this: https://www.facebook.com/GoldenCrest...162542845/?t=0

A lot of indecision by me, and them Faulkner bails me out with a great play.

I'm scoreboard-watching the Twisters and Swarm, and I'm starting to think this isn't going to be the Treefrogs logo year.

Seriously, 49-0? 41-10?

My 28-6 game felt like total domination. I can't imagine how the Swarm and Twisters games went.

[GoldenCrest Games]

Fast-Sim Update:

Kickoff: DONE
Passing game: fully functional, not balanced
Running game: fully functional, not balanced
Interception returns: fully functional, not balanced
Open field running: fully functional, not balanced
Punting: DONE
FG: DONE
Integration into existing coach/playcall AI: DONE
Stat tracking: 1% functional
Integration into existing schedule screens: 0% functional

It's really going to be a relief when this is done. This is one of those lessons-learned about planning out what you're doing before you dive into it. (Granted, I'd never intended on having a fast-sim, but still....)

The most complicated part of fast-sim has been all the helper functions and coach AI that were built with the knowledge that they could just reach into the game engine and pull out game variables whenever needed. Unfortunately, fast-sim is not the same as the original game engine, so all of those reference won't work anymore. It's incredibly tedious trying to unwind all that mess, and make it work for two types of sim engines, without creating duplicates of everything.

Now, if I was smart and had planned all this out from the beginning......

1. I would have made a single point of entry to both game engines, so in-game variables were only stored once.
2. I would have written AI so it didn't rely so heavily on being able to peek into the game engine at will.
3. I wouldn't have used GameObjects AT ALL for tracking players (they are the worst to work with)

Those simple changes would have made this whole process easier, and saved so much time. Just one more thing to remember for the golf game.

[jct32]

Wow 49-0? Seems too good to be true.

[GoldenCrest Games]

Originally Posted by jct32

Wow 49-0? Seems too good to be true.

That's a scary team now.

[GoldenCrest Games]

Week 5

I finally got a taste of Swarm/Twisters level domination.

This week I switched over to the Swarm offensive playbook. As I was playing with my defensive depth chart, it dawned on me that my new CB Benn is a pretty effective RB/WR. (maybe even more effective than Faulkner!) I'm happy with the Spider playbook now, so I switched over to the Swarm power-passing offense.

O.M.G.

That is an INSANE offense. I can see why the Swarm are so powerful. If you have the RBs to make it work, and have at least 1 legit WR threat, it's an offense that is really hard to stop. I'll need to run it a few more games to figure out precisely why it's so tough, but right now it feels like the delay for the RBs to get upfield and into pass patterns is creating extra space for them to get open. I do't think it's the matchups, because the defense always matches up man defenders based on receiving threat.

In any case, it's a really cool offense. I'm loving it. Also, Benn is good enough that Faulkner might permanently have a new running mate.

I captured a little video of Faulkner working over the Alpacas on a kick return. Just look at that move!!https://www.facebook.com/GoldenCrest...341884627/?t=0

I'll try to get some good videos of the power-passing offense in action.

[GoldenCrest Games]

I'm not too surprised to see the Swarm, Twisters, and Treefrogs at the top. The Swarm and Twisters coaches have been fine-tuned to their playstyle, so the AI does a great job of managing the teams. The Treefrogs are always a regular season hero, but a postseason zero.

What's cool is how the Maroons are doing! Remember, they traded away the #1 pick this offseason. It sure seemed to work out for them, because they are rolling pretty well for a team that had 1 win last season. (and not a lot happened in FA this year)

The Blues, on the other hand, kept their pick and look about as you'd expect from a team building through the draft. Incremental improvement.

The new MVP scoring system is definitely shaking things up this season. Armbruster has taken his place at the top of the rankings, which is really where he should be. He's the most talented player in the league for sure, and now he's getting recognized by the media for it. Also, how about his QB getting him the ball now? #2 on the receiving yardage chart (at the WR position), from a team that almost exclusively throws to the RBs? Wow.

Luna, as always, is making a mockery out of the passing leaders board. Come on man, save some passing yards for the rest of us!

[GoldenCrest Games]

Whoah! They just unblocked FOFC from work!

[jct32]

That cut inside by Faulkner was very nice.

Looking forward to the next week!

[GoldenCrest Games]

Week 6

The game against the Avocados was pretty uninteresting. They don't have a meaningful QB so they didn't put up much of a fight. The only reason they scored at all is because I started going for it every 4th down.

That aside, this also happened: https://www.facebook.com/GoldenCrest...284494366/?t=0

WTF is that linebacker doing??? It's been a long time since I've seen this happen. I guess there's a little more AI work to do......

Twisters and Swarm roll again. I'm starting to feel like I have a chance, though. The power-passing offense is much more effective for my team, and my defense is significantly better than before. I've moved Buckles back to S so now I have a legit player at every level. That really helped against big plays. He usually makes the tackle before anything is broken long.

Probably the first time I've said anything nice about the S position.

[GoldenCrest Games]

Standings

I can't get over the passing advantage the Twisters seem to have over the rest of the league.

Luna is only rated an 84
They don't have a WR with a speed greater than 79
WR3 has a speed of 59

Also, there are 5 other QBs within a couple points of Luna's rating, and some of those guys have significantly better receiving weapons.

If he even showed up on defense or special teams, he'd own the MVP this year.

[GoldenCrest Games]

Profiles for the big-3

• Kory "I ain't got time for defense" Luna
• Reginald "Oh sure, NOW you notice me" Armbruster
• Tyree "Momma, I'm goin' fast!" Faulkner

[GoldenCrest Games]

Fast-Sim Update:

• Kickoff: DONE
• Passing game: DONE
• Running game: DONE
• Interception returns: DONE
• Open field running: DONE
• Punting: DONE
• FG: DONE
• Integration into existing coach/playcall AI: DONE
• Stat tracking: 1% functional
• Integration into existing schedule screens: 0% functional

Basically every phase of fast-sim works now. It's hard to say how far off the stats are from the real-time engine, but there's a TON of overlap so I bet it's not too far off.

Tomorrow I should be able to start implementing stat tracking. That's when the fun begins! After stat tracking is done, I'll run off this season a bunch of times to get a fast-sim result to compare against the current season progress.

That should be... interesting.

[GoldenCrest Games]

For those who are interested in release plans, check out the blog. http://www.goldencrestgames.com/fast...release-plans/

[jct32]

The system works. Wait until we get some quality players in there.

I also find it interesting that the defense can add for the MVP when Luna has a 29:1 TD/INT ratio. Then I realize that this is slightly different when compared to real football. The return game is stronger here and more important. If I quit comparing this to real life football (like having 2 way players) then everything seems a lot better and your leagues seems like it is really humming.

[GoldenCrest Games]

Originally Posted by jct32

I also find it interesting that the defense can add for the MVP when Luna has a 29:1 TD/INT ratio. Then I realize that this is slightly different when compared to real football. The return game is stronger here and more important. If I quit comparing this to real life football (like having 2 way players) then everything seems a lot better and your leagues seems like it is really humming.

I have a hard time with that, too. I often catch myself thinking of the stats in terns of the NFL, and then think I need to fix something. I have to keep telling myself that it's totally different, and it's ok if the stats don't match up.

That being said, I think some of the defensive and return stats might be a bit over-valued right now. After each development season, I'll run the season stats through my MVP scoring system and rebalance the weights. This season's scoring system is based only off last year's stats, so it's probably not quite right because of how the engine changed during last season. The next balancing should get it a lot closer.

[jct32]

I think that will be one of the biggest hurdles will be people expecting this to be like NFL or NCAA. Will you allow people to adjust modifiers to adjust the offense, defense or special teams?

If you allow people to adjust the league to play they want then I think you have a really solid game. However, if you can’t then I think it might take a little longer for people to adjust to this version of football.

[GoldenCrest Games]

Originally Posted by jct32

I think that will be one of the biggest hurdles will be people expecting this to be like NFL or NCAA. Will you allow people to adjust modifiers to adjust the offense, defense or special teams?

If you allow people to adjust the league to play they want then I think you have a really solid game. However, if you can’t then I think it might take a little longer for people to adjust to this version of football.

That's one of the main things I want to get from the Alpha testing phase. Are people able to quickly get into it, or does the difference in stats turn out to be a barrier?

[GoldenCrest Games]

On a related note, I almost made this first version a Blood Bowl or Mutant League Football type game, just to force the idea that it's different than the NFL.

[SirBlurton]

Originally Posted by GoldenCrest Games

On a related note, I almost made this first version a Blood Bowl or Mutant League Football type game, just to force the idea that it's different than the NFL.

Glad you didn't because that's one of the ideas on my "someday I'll make" list!

[GoldenCrest Games]

I also had a Cyberball remake planned before this game, but had a hard time finding the right models. I think if I'd have found the robot models before I found the football player model, things might have gone a lot differently.

[GoldenCrest Games]

A lot of close games this week.

How about the Swarm and the Twisters winning by a combined 5 points?? That's unbelievable.

Our game against the Silvers wasn't as close as the score looks. My playcalling was pretty sloppy most of the game, and went for it on every 4th down. The Silvers had nothing on offense, and I could sack the QB at will. I called off the dogs early and started messing around with some defenses I rarely run. They broke through for a long TD late in the game, but I wasn't too worried.

I had a problem running the ball this game because they have an elite 2-man DL that blew me up consistently. It's hard to run when 2 guys destroy your blocking and leave all the other defenders free for tackling.

I ended up relying completely on the power passing attack. Here's a quick video of my new favorite play: https://www.facebook.com/GoldenCrest...623905132/?t=0

It almost always gives the QB at least one good passing option.

1. The WR1 on the hard slant&drag is the first option. If there's a mismatch in man defense, the QB usually has a good angle to get him the ball. If it's a zone, the RB clears out the middle.
2. If the WR is covered, or if the zone stays intact, the deep RB (RB1) is the next read. The lead RB (RB2) will usually occupy the LB enough for RB1 to get wide and open. In this video, the LB did a good job of coming off RB2 to cover RB1 and take it away.
3. RB2 is the 3rd read. A lot of times if WR1 and RB1 are covered, it's because the defense is playing shallow and loading up in the first 10 yards. RB2 is challenging the defense deep, and will often get a step on the safety if he starts too close to the LOS.

My QB is pretty quick at making reads, so I think he threw the ball on his 2nd cycle of reading the receivers. It looks like everyone was covered on the first reads, and then Moody found the WR at the beginning of his 2nd read cycle. He's rated 80, and a 70 or 60 rated QB probably would have been looking at the RB1 when the WR found the hole over the middle, and would have missed it.

That's really the most meaningful difference in QB skill. It's the ability to read receivers quickly. Bad QBs also are inaccurate with their throws, but the accuracy penalty isn't really bad unless you're throwing 20+ yards downfield. Being able to cycle through receivers quickly is key because they are usually only open for a small window. Most defenses recover quickly, unless it's just a bad play design. (there are still some of those)

If you've got a crappy QB, you're going to be limited to patterns with 1 or 2 receivers, or you'll just have to live with only the top reads in complex patterns. It's doable, but it's really limiting.

Probably my favorite feature in the game.

[GoldenCrest Games]

The first ever complete fast-sim game done!

Twisters vs Treefrogs

Final score: Twisters - 3, Treefrogs - 7
Treefrogs scored on the final play of the game, with a 86 yard TD run!

ok, that's not totally out of the realm of possibility. I'm pretty sure there's a ton of tuning needed, but it's a good starting spot. Now for some stats:

Passing
Luna: 15/23 for 117 yds (not horribly out of line. Probably a bit low)
Moody: 7/15 for 36 yds (same as Luna, seems a bit low)

Rushing
Twisters: yeah right.
Faulkner: 2 att for 90 yds. (my AI coach never wants to run. This might work itself out with more attempts.)

Receiving
Horowitz (Twisters): 9 rec on 14 att, for 54 yds. (The yds per catch is low, when compared to the real-time engine. Not horrible, though)
Karp (Twisters) 5 of 5 for 59 yds. (not bad)
Faulkner (Treefrogs) 4 of 8 for 17 yds. (yeah, yds per rec is low across the board)
Buckles (Treefrogs) 3 of 5 for 18 yds

Kick Return - seems ok for now
Faulkner 2 for 43
None for the Twisters

Punt Returns - overall looking ok
Faulkner 2 for 79
Cromwell 3 for 115

Defense:
0 sacks for the game. That doesn't seem right, but both QBs have a pretty quick release. Might need watching.
0 ints. Again, might need watching but this isn't unusual for 80+ QBs
Tackles seem to be distributed realistically. CBs got most of the tackles, which is expected because we were both passing all game.


Right now it looks like my open field running is too hard on the runners. Receivers aren't getting and RAC and returners are being held in check. I'm actually a little shocked about how good the stats look for the first run. I was expecting something like a 99-80 score or Luna getting 700 yards passing.

[jct32]

Grabbing the pitchforks!!!

3 points!!!

DOWN WITH MANAGEMENT, DOWN WITH THE COACH, DOWN WITH FAST-SIM!

Joking aside, it does seem like the engine is slightly off. I don't think you would get that same result if you played the game.

[GoldenCrest Games]

Week 8

We got to play the no-pass Blues this week. It was pretty interesting to play a team that is allergic to throwing the ball. They often mounted long drives, but eventually stalled out when they hit a 3rd and 4+ yards. Often they went for it on 4th and short, and often got it.

The run a stout OL with a incredibly skilled, but slow, RB. Here's a video of a few Blues plays, ending with an atrocious pass attempt.

https://www.facebook.com/GoldenCrest...080317653/?t=0

Watch how quickly the RB cuts. That's what you get from high RB skill. You get quick decision making. If you go back and watch any of the Faulkner runs, you'll see his decision making is slightly slower, and he'll run straight-line longer before making another move. On the other hand, Faulkner in kick return makes lightning quick cuts. That's because he's very skilled as a KR.

For the Blues, this means a LOT of 4-6 yard runs. He almost always makes the most out of his blocking, but isn't fast enough to break the home-run. Pure ball control.

We held out own on offense. We were solid, but unspectacular. Faulkner did have his first punt return for TD, though. That was pretty cool.

[GoldenCrest Games]

He might not be at the top of the MVP race, but Horowitz is having a beast of a year! The new MVP ratings system is really highlighting the 2-way and 3-way players.

Just over halfway through the season, and it's pretty clear the Twisters, Swarm, and Treefrogs are in the playoffs. The Oranges look legit, but everyone else looks pretty bad.

Also, let's not forget that our league rushing leader plays QB. That could be dangerous to play against in the playoffs.

[GoldenCrest Games]

The first full-week fast-sim test is in!

Old sim time: 40-45 minutes
Fast-sim time: 8 seconds

No bad.

[GoldenCrest Games]

I'm fast-simming the current season up to where we are now, and it's taking just over a minute. Sweet!

I've also been tuning the fast-sim and the stats are coming out pretty close to what we have so far this season. If anything, they actually look a little more realistic because there's not such a huge distance between the #1 guy in a category and everyone else. I still have some work to do on sacks and ints, but it's close. (I'm hesitant to do too much with sacks and ints because I want to work on those in the real-time engine, too.)

I'm messing with some ways of posting the fast-sim results against the real league results that won't be confusing. I don't want the real league results to get confused with testing!

The plan is to finish this season out in the real-time engine, and starting next season we'll go with the fast-sim for all non-human games. (playoffs tbd...)

[GoldenCrest Games]

It kind of bugs me that my coach thinks we're a passing team. Every time I let the coach manage the game for me, we always go all-pass.

I get that Faulkner is only rated 70 at RB, but come on!

(this might be another instance where the AI is better at this game than I am)

[jct32]

Great to hear the development on this! Looking forward to more updates!

[GoldenCrest Games]

Yesterday I collected 5 seasons worth of test data to compare against the real sim data, and found a few areas that need some more work.

1. Running looks like it's close, but when you dig a little you see that the yardage is almost all coming from breakaway runs. There are very few 4-10 yard runs. I think that has to do with the blocking system, and should be fairly easy to tune.
   
2. Passing is really close, but there aren't enough breakaway receptions. I made a small change to the running system that resulted in all those breakaway runs, so I think I can apply that (to a smaller extent) to the post-reception running situation.
   
3. I also realized that digging in the data takes too much time. After exporting it, I have to pull it into my master spreadsheet and do some pivot tabling to it, then set up a way to view it easily. I'm kind of lazy, so that's too much work to do every time I want to run a test season. If there only was an in-game screen that could show me the stats for any category I needed.......

Yeah, It's finally time for the league leaders screen. It's going to be a huge help to view stats in-game, without having to do all the exporting/importing/formatting every time.

[GoldenCrest Games]

Well, now we know. It's not just the fast-sim that needs more work.

Here are the season passing leaders for the league. I'd like to see a more gradual falloff from the top QBs to the middle/bottom QBs. That being said, I know the running (and returner) teams will have a massive gap at the top as well. It's probably a result of how specialized everyone is, which might be ok.

I'd still like a more robust middle-class, though.

[GoldenCrest Games]

Originally Posted by GoldenCrest Games

I can't get over the passing advantage the Twisters seem to have over the rest of the league.

Luna is only rated an 84
They don't have a WR with a speed greater than 79
WR3 has a speed of 59

Remember this oldie?

Well, here's why Luna is dominating the league. It's Horowitz.

75 receptions in 8 weeks??? It's like he gets open EVERY PLAY!

[GoldenCrest Games]

Week 9

This was a rough one. We got thoroughly worked by the Swarm. The only spark of life was a kick return for TD after we were already down 14 points. I'm not sure we even got a single first down the whole game. That defense is just smothering.

Game highlights:https://www.facebook.com/GoldenCrest...076307520/?t=0

How about the Twisters losing to the Olives?? That was unexpected. Also, the Oranges (who are the only other meaningful team in the West) getting beat by the Blues. Crazy day.

(note: The real-time engine code hasn't been touched this season.)

Development notes:
Fast-sim passing and rushing stats are starting to come into line with what I'm looking for now. I need to work on sacks/defense and the return game next. Sacks are super rare, punt returns are ridiculously high, and kick returns are super low. Everything else is looking pretty good.

After fast-sim is done, I'll have to tune the real-time engine to match. Now that the leaders screen is working, I can see that a lot more tuning needs to be done on the real-time engine.

[jct32]

Got beat by the Olives? Well they are the second best in our conference. Can't be too upset about that. Just gotta pick it up next season.

Good stuff on fast sim. I think you will correct those issues quickly.

[GoldenCrest Games]

I'm down to 2 remaining issues with fast-sim:

1. There are no kick return TDs
2. Punt return average way too high, but there are also no punt return TDs.

Everything else (including playcalling) looks pretty darn good now. Tomorrow morning I think I'll be able to get the KR and PR issue fixed, and then it's just a matter of tuning the real-time engine to match.

Still looking good for an August alpha release!

[GoldenCrest Games]

And just like that, Fast-Sim is done!!!

There will be some balancing negotiation between the engines as the test seasons play out, but I'm ready to call the engine done and move on to the next thing.

That's quite a relief! I started getting so bored with the engine that I almost scrapped the whole thing and started re-writing the game to work in real-time only!

[jct32]

Congrats on getting fast sim done!

I'm excited each step this progresses.

[GoldenCrest Games]

Week 10

The Olives played us tough for most of the game. We had a hard time getting Faulkner any space, and the rest of the WRs weren't catching anything. Just a hard offensive outing.

We had one nifty power-pass play that went for a long TD, but then it quickly got erased. (video here: https://www.facebook.com/GoldenCrest...297262831/?t=0 )

Ugh!!!

I'm loving the new-look defense. My CBs aren't all-world, but I can count on them to provide decent coverage and to pick off a pass now and then. I can live with that. It sucks that I had to give up the all-time great OL, but it's probably worth it in the long run. It's just frustrating giving up sacks and not being able to run the ball anymore.

In other news, the Twisters and Swarm finally get that rematch from the Championship game! AND END IN A TIE!??!?!!? LOL!!!!!
That probably speaks to the Twisters stepping up the defense this year. Should be a good finals again.

p.s. I'm not happy to see the Nemesis get new life.

[GoldenCrest Games]

In developer news: With fast-sim done, I'm turning all my attention towards getting a playable version out for you guys to mess with.

Here's the list as I see it now:

• Team selection screen
• Start league with allocation draft
• Clean up save/load
• Make sure league phases transition cleanly (just some minor screen variable cleanup)

That's pretty much it! If I can fix all that, we should be able to have something to send out!

[jct32]

Wow, I wouldn't expect a tie. Also don't forget your own team. You have been doing really well with that power offense.

Once again thanks for the update on the development. Sounds great!

[GoldenCrest Games]

Originally Posted by jct32

Wow, I wouldn't expect a tie. Also don't forget your own team. You have been doing really well with that power offense.

Love the power-passing offense! I'm going to miss it when I take on a new one next season.

I'm already worried about another quick exit from the playoffs. We don't exactly have the best track record.

[GoldenCrest Games]

We had another tough offensive game in week 11. For the most part, the Cyans held us in check. There was absolutely nothing to be had in the running game. It pretty much went like this the whole game:
https://www.facebook.com/GoldenCrest...060412988/?t=0

It's becoming a pattern now, where our only offense is coming from Faulkner getting loose on a pass pattern and Moody hitting him for a long TD. Nobody else on my offense is providing anything at all. WRs are constantly dropping long passes, Moody is always getting sacked, and my OL can't make a running hole to save their life.

But I still have Faulkner.

Thank goodness my defense is still killing it.

The Twisters grab another tie against the T-Pain. I imagine they are just trying to make the rest of the league feel better about themselves.

Swarm beat up on the Eggshells, as expected.

Also, the Nemesis lose. Nice!

[GoldenCrest Games]

Let's take a quick look at the standings, with 3 games to go in the regular season.

In the West
Swarm and Treefrogs have clinched. It looks like it's going to be the T-Pain and Oranges joining us. That last loss probably puts the Nemesis out of the playoff picture.

Good.

In the East
The Twisters have clinched, as expected. I'm positive the Rainbow Sharks and Olives are in. If I were to lay money down on the 4th slot, I'd actually pick the Maroons to make it into the playoffs over the Eggshells and Slates. I really like their point differential and the defense. The Maroons also have loaded up on special teams, which doesn't always show up in the stats very well. That's my darkhorse pick!

Now that I've gotten the fast-sim done, and have been looking at the stats a LOT, our league stats just feel out of whack to me. This is probably the last time you see the league leader in passing (and receiving) almost double-up the 2nd place QB. Yikes!

Come on, Luna! Leave some passing yards for the rest of us!!

[jct32]

Nerfing my offense again? I can't catch a break can I?

[GoldenCrest Games]

Originally Posted by jct32

Nerfing my offense again? I can't catch a break can I?

Ha! Just in time, too!

As luck would have it, it was Treefrogs-Twisters week!

The Twisters defense has clearly been upgraded. My passing game now knows that for sure. It was a pathetic offensive showing for us, but Faulkner put us on his back and carried us through the game. Frankly, he's becoming a legend around these parts. I don't think our team would have any relevance without him. He may be gaining lifetime-contract status.

The Twisters offense is really hard to stop. They aren't fast or flashy, but they complete a lot of passes. The thing that separates the Twisters passing attack from everyone else, is that they catch almost everything. I can get guys open at about the same rate as the Twisters, but my guys can't catch in traffic. If they have a tiny amount of space, the Twisters WRs always catch the ball.

That being said, Luna made one mistake that led to an interception while they were in FG range. That was the difference that allowed us to hang on for the tie.

Yep, another Twisters tie!

Here are the scoring highlights: https://www.facebook.com/GoldenCrest...163348411/?t=4

In other news, the Swarm grab another win and probably lock up the #1 seed. The Eggshells and Slates both lose while the Maroons win! I just had a feeling about them. I'm really rooting for them to make it to the playoffs. It would be cool to see the worst team from last year make the turnaround.

[GoldenCrest Games]

Look! Team Logos are starting to make an appearance!!



I'm just starting to incorporate logos right now. Most of the graphics work is going to take place in Dev Season 4 or 5.