TBCB Random Fighter Generator v3.2 released

[PyroFalkon]

v3.2 (23 December 2004)
Decreased the chance that Clinching will be less than 62.

I really hate releasing two versions so close to each other, but in my standard 5 boxer field test every time I make a change, I noticed that Clinching was all over the place. I'm actually surprised I hadn't noticed it before.

Anyway, if you just got 3.1, you PROBABLY don't need this. Unless you like downloading the latest version, or the wacky Clinching numbers really bother you.

Oh, you'll also notice a change of the name of the ZIP file. Nothing big, but this way the version number is attached to the ZIP for easier reading or something.

[Bonedwarf]

Funny. I was going to suggest changing the filename yesterday

Good work!

[PyroFalkon]

bump

[PyroFalkon]

bump again. Can I get a sticky up in here?

[Bonedwarf]

Come on lads, give the man some glue!

[Rondo_GE]

Originally Posted by PyroFalkon

v3.2 (23 December 2004)
Decreased the chance that Clinching will be less than 62.

I really hate releasing two versions so close to each other, but in my standard 5 boxer field test every time I make a change, I noticed that Clinching was all over the place. I'm actually surprised I hadn't noticed it before.

Anyway, if you just got 3.1, you PROBABLY don't need this. Unless you like downloading the latest version, or the wacky Clinching numbers really bother you.

Oh, you'll also notice a change of the name of the ZIP file. Nothing big, but this way the version number is attached to the ZIP for easier reading or something.

Outstanding!

You might want to check the "Hitting Power" ratio. On first try I did about fifteen fighters with it and came up with three guys with 12 ratings. One of the 12 rated guys had a Go For Knockout strategy rating of "0"...well thank god for that!

There were also no 7 and 8 rated fighters in the group who I think would be the most frequent.

I did another group (of 25 this time) and came up with 3 13's rated like that and still no 7 and 8's.

Seems like you need a bell curve of some sorts here. I used to be able to randomly simulate dice rolling in Basic but I don't remember the command something like ((rnd(8)+1)+(rnd(8)+1)) (for an 8 sided die). What's the upper and lower limit on hitting power?

If I use these fighters it's gonna be one wild division!

[PyroFalkon]

The biggest problem I have with the RFG is that it's TOO random. I need a way to control it (like a bell curve), but it's beyond my logical brain. Chris is back though, so I'll work with him on it.

As I said, my "dream" is to allow you to choose the overall rating and create a guy that way, but because Andreas won't release the algorithims to how the overall rating is determined (which I totally understand and respect), I'm not sure how I could go about it.

But as I said, Chris is back from the Air Force, so we'll figure something out. I won't give a date or anything for the next release because I don't have a date myself, but just know I haven't quit or anything.

[dlirag]

Can you take us through the program's random formula for HP again?

[PyroFalkon]

Hitting Power is randomly chosen from 0 to 13 without a weight. (That means, there's an equal chance of any number being chosen.) There's roughly a 1:3300 chance that you'll see three 12s out of 15, so you may not see it again. I think it was just coincidence.

Hitting Power DOES correlate to the 3-point punch ratio. In one page of the manual, it says what the total percentage should be for 3-point punches. For 12 for example, it's 45-47. I have my program assign points randomly to the five punch types in 3-point punches first, and then it stops when it gets in the 45-47 range (the SPECIFIC limit is randomly chosen too, so it veries from fighter to fighter). Then it assigns the rest to two-point punches.

HP does not correlate to anything else, however. I suppose it SHOULD go to Go For KO or something, but again I'm a boxing newbie and don't QUITE know how things work. I read in the manual somewhere that a particular boxer killed someone in the ring, and after that could never really KO someone. So his Hitting Power might be high, but his Killer Instinct and Go For KO should be really low.

The point is, anomolies happen. However, if they're happening TOO frequently, then I'll put a weight on it. Bell curve or something.

[Johnny Slick]

I don't know what system you're using for randomization, but why not try 4d4-3? That's coming up with a random number from one to 4 4 separate times, adding up the total, and subtracting 3. You'd end up with a curve heavily weighted towards the middle, with just a few 13s and 1s (1 out of every 256 boxers for each).

[PyroFalkon]

What is the math for that? I know about the dice rolls (used to be a big D&D player), and something in my brain told me that doing it that way would make better results... Could you show me the reason the dice roll system is better than just pure randomization?

[Johnny Slick]

Originally Posted by PyroFalkon

What is the math for that? I know about the dice rolls (used to be a big D&D player), and something in my brain told me that doing it that way would make better results... Could you show me the reason the dice roll system is better than just pure randomization?

Well... it IS pure randomization. It's just that having multiple "dice" creates a natural bell curve. Think about it this way: If you have 2 6-sided dice, the only way you're going to come up with a 12 is if both dice come up 6es. Here's a representation of the possible outcomes:

XX| 01 02 03 04 05 06
----------------------
01| 02 03 04 05 06 07
02| 03 04 05 06 07 08
03| 04 05 06 07 08 09
04| 05 06 07 08 09 10
05| 06 07 08 09 10 11
06| 07 08 09 10 11 12

7 comes up 6 times as often as 2 or 12. If you add in more dice, the bell gets even taller. I'm sure there's a quick and dirty way of figuring out the chances of getting a 10 from 4d4, but I don't know of any way to do it other than to just write out the totals...

2d4
xx| 01 02 03 04
----------------
01| 02 03 04 05
02| 03 04 05 06
03| 04 05 06 07
04| 05 06 07 08

2 1
3 2
4 3
5 4
6 3
7 2
8 1

Okay, so figuring out the math here...

2 - 4(1), 5(2), 6(3), 7(4), 8(3), 9(2), 10(1)
3 - 5(2), 6(4), 7(6), 8(8), 9(6), 10(4), 11(2)
4 - 6(3), 7(6), 8(9), 9(12), 10(9), 11(6), 12(3)
5 - 7(4), 8(8), 9(12), 10(16), 11(12), 12(8), 13(4)
6 - 8(3), 9(6) 10(9), 11(12), 12(9), 13(6), 14(3)
7 - 9(2), 10(4), 11(6), 12(8), 13(6), 14(4), 15(2)
8 - 10(1), 11(2), 12(3), 13(4), 14(3), 15(2), 16(1)

4 - 1 (0.3%)
5 - 4 (1.6%)
6 - 10 (3.9%)
7- 20 (6.8%)
8 - 31 (12.1%)
9 - 40 (15.6%)
10- 54 (21.1%)
11- 40 (15.6%)
12- 31 (12.1%)
13- 20 (6.8%)
14- 10 (3.9%)
15- 4 (1.6%)
16- 1 (0.3%)

[PyroFalkon]

Okay, you know what? My favorite gambling game is craps, because of how the odds are almost on your side. It all revolves around the number 7, because 7 comes up more times with two dice than any other number. You'd think that because of how much I've studied craps, I'd remember the bell curve.

You ever have one of those moments when you wanted to remove your brain, dropkick it, and say "Why in the bleeping bleep bleep didn't I think of that?" (Except I actually SAY "bleep" instead of naughty words, because I'm different.)

Huh. You've given me a ton to think about, Slick. As it is, I don't think I should use a perfect bell, because then the lowest end of the scale has just as much chance as someone at the high end of the scale. I think if I incorporate a bell curve, a perfectly good score should be LESSER of a chance than a perfectly bad score. I wonder how I need to put THAT into math...

But that's stuff Chris and I need to deal with, next time he gets his butt over here. Well, at least I know I won't be bored this week. Thanks a million, Slick! The RFG can only get better.

[Rondo_GE]

Originally Posted by PyroFalkon

The biggest problem I have with the RFG is that it's TOO random. I need a way to control it (like a bell curve), but it's beyond my logical brain. Chris is back though, so I'll work with him on it.

As I said, my "dream" is to allow you to choose the overall rating and create a guy that way, but because Andreas won't release the algorithims to how the overall rating is determined (which I totally understand and respect), I'm not sure how I could go about it.

But as I said, Chris is back from the Air Force, so we'll figure something out. I won't give a date or anything for the next release because I don't have a date myself, but just know I haven't quit or anything.

You can simulate a dice roll (at least two die) or.....

You might want to think of something else....One thing I would do is actually look at the TCB database of existing fighters for each weight class and find out what the real percetages and frequency are and then see if you can come up with an equivalent curve....Remember it is probably more important to use non retired group rather than a retired group.

So let's say you look at 200 heavyweights and this is what you come up with...

HP
13--(0)
12--(0)
11--(1)
10--(12)
9--(27)
8---(71)
7--(52)
6--(23)
5--(12)
4--(2)
3--(0)
2--(0)
1--(0)

The breaks out as....

13--0
12--0
11--0.5%
10--6%
9--13.5%
8--35.5%
7--26%
6--11.5%
5--6%
4--1%
3--0
2--0
1--0

Now you could go with these numbers (not recommended) or you could "smooth the curve" by doing this....

Find the median and then average each corresponding number on either side of the distribution (for example 7 and 8 percentages average out to 19.75) so the above curves smooth out to…

13--0
12--0
11--0.75%
10--6%
9—12.5
8—30.75
7—30.75
6—12.5
5--6%
4--.75%
3--0
2--0
1—0

which I would round to something like this…

13—1 die roll(100)
12—1 die roll(99)
11—2 die roll (97-98)
10—6 die roll(91-96)
9—11 die roll(81-91)
8—29 die roll( 52-80)
7—29 die roll (23-51)
6—11 die roll (12-22)
5—6 die roll (6-11)
4—2 die roll ( 4-5)
3—1 die roll (3)
2—1 die roll (2)
1—1 die roll (1)

If this looks like work I guess it is. I think things would be easier if you could extract a range of fighters from the program in excel, access, or some kind of CSV file. Of course you could just bypass any analysis and do what we call a reasonable “swag” but even there I would try to come up with an accurate percentage of the top three most frequently used HP ratings so you know where to place the high end of the bell.

Also bear in mind it might be advisable to skew the bell up and down one notch for three general categories…

+1 for Light heavyweights, Cruiserweights and Heavyweights…

No nudge for Lightweights to Middleweights

…and –1 for all classes below lightweight.

…or just leave it to the user to use common sense in that regard.

Sorry about the long post.

[PyroFalkon]

Rondo --

First of all, I apologize for the long delay in replying. I must have gotten the e-mail that this thread got a new post, but then lost my mind and forgot to read it.

Second, I hate it when people apologize for the length of their posts, e-mails, and everything else. The RFG practically RELIES on your input (and I'm talking to every single poster here). Rememer, Chris and I know NOTHING about boxing. Hell, I didn't know the difference between a jab and a straight punch until I played Fight Night 2004, and I *still* don't know how to recognize a cross. The point is, we rely on all of you to help us figure out the nuances to boxing, so any input you have is valuable.

And your post, Rondo, was quite valuable. You're right that it's work to go through 3000 boxers and find trends, but realize this: if I wasn't willing to work to make the RFG fantastic, I would have just left the damn thing on version 1.0 and let everyone deal with a subpar utility.

As it is, I love the RFG and want to make it perfect, no matter how much work that takes. And to be honest, I have looked into numbers and so on, but I haven't publicized it. Chris and I have actually had a minor meeting, and we're planning on not only fixing the RFG for a natural bell curve, but to create a new utility that will make almost everyone here wet themselves with excitement.

However, no more details about THAT will be forthcoming. I'm not trying to keep it a secret intentionally... We just don't know yet whether the new utility will actually work, and if it doesn't, I don't want to disappoint everyone here by having made an empty promise.

That said, one detail I *will* spill is that we have solved the bell curve thing. In fact, over this weekend, great strides will be made on the RFG, and a new version SHOULD be out early next week. Unlike previous versions however, I'm keeping the old v3.2 around in case the change turns out to mess with things too badly.

Now, the next version will probably get the number v4.a, because I'm 98% sure it's going to be incomplete. It's basically an "unfinished" released, and we will need an equivalent of beta testers to make sure we didn't do anything completely off the wall. That's the other main reason we're keeping v3.2 around; it's going the only "finished" release until we're ready to the stamp of approval on v4.0.

If anything changes, I'll be sure to put a new thread. Until then, keep on flinging suggestions as long as often as you want. Trust me, I read them all.

[Rondo_GE]

Originally Posted by PyroFalkon

Rondo --

First of all, I apologize for the long delay in replying. I must have gotten the e-mail that this thread got a new post, but then lost my mind and forgot to read it.

Second, I hate it when people apologize for the length of their posts, e-mails, and everything else. The RFG practically RELIES on your input (and I'm talking to every single poster here). Rememer, Chris and I know NOTHING about boxing. Hell, I didn't know the difference between a jab and a straight punch until I played Fight Night 2004, and I *still* don't know how to recognize a cross. The point is, we rely on all of you to help us figure out the nuances to boxing, so any input you have is valuable.

And your post, Rondo, was quite valuable. You're right that it's work to go through 3000 boxers and find trends, but realize this: if I wasn't willing to work to make the RFG fantastic, I would have just left the damn thing on version 1.0 and let everyone deal with a subpar utility.

As it is, I love the RFG and want to make it perfect, no matter how much work that takes. And to be honest, I have looked into numbers and so on, but I haven't publicized it. Chris and I have actually had a minor meeting, and we're planning on not only fixing the RFG for a natural bell curve, but to create a new utility that will make almost everyone here wet themselves with excitement.

However, no more details about THAT will be forthcoming. I'm not trying to keep it a secret intentionally... We just don't know yet whether the new utility will actually work, and if it doesn't, I don't want to disappoint everyone here by having made an empty promise.

That said, one detail I *will* spill is that we have solved the bell curve thing. In fact, over this weekend, great strides will be made on the RFG, and a new version SHOULD be out early next week. Unlike previous versions however, I'm keeping the old v3.2 around in case the change turns out to mess with things too badly.

Now, the next version will probably get the number v4.a, because I'm 98% sure it's going to be incomplete. It's basically an "unfinished" released, and we will need an equivalent of beta testers to make sure we didn't do anything completely off the wall. That's the other main reason we're keeping v3.2 around; it's going the only "finished" release until we're ready to the stamp of approval on v4.0.

If anything changes, I'll be sure to put a new thread. Until then, keep on flinging suggestions as long as often as you want. Trust me, I read them all.

Well if you keep working it the way you are I think you'll get it down pretty good.

Once you figure out how to create bell curves I think you are on your way to having this thing licked.

Another thing to bear in mind, eventually, is that size and weight effect some divisions more than others such as the HW divison where you have different kinds of body types. I guy with a long reach is more likely to have a better and more frequently used jab than a guy with a short reach and is in fact short who might prefer hooks and uppercuts or punches in combo. Size and weight would also effect the way a fighter's dominant reliance on strategies goes (to some degree)...tall lanky guys would be more likely to be efficient at gith outside both offensively and defensively than shorter guys.

So you might want to generate some physical stats to begin with. But that's in the future I think.