Game Analyzer

I am nearing completion of the game stats accumulator part of the game analyzer. Here is a partial list of the accumulators i am keeping

Game:
map, start/end, pills captured, builders lost, tank kills, and base captures, plus a couple more things.

Player:
games played, tanks killed, tanks lost, bases captured, bases lost, pills catured, pills lost, builders lost, and a couple more things.

I have assigned an internal id for both the game and player, and condense the game log to game, action, player1, and player2 (where player2 is the person who lost the tank, pill, base)

After I get the initial cut done, i'll post a page that has the game averages, and where each player is in relationship to the overall average. I found this informative when just looking at myself and how my games stack up against other players in the same game.

Just wanted to let everyone know where i was at.

niiiice, can't wait to c it when it's done!

I've got the accumulators working fairly well, i've posted some initial player and game averages at:

Player Average
Game Averages

I'm moving into smoothing the data, and generating the ratios and stuff for the standard profiles... should take a couple more days to load 270 more games.

requesting representation on the stats list.

Xenocide wrote:

requesting representation on the stats list.

I ran most of the games from today thru, never saw your name... sure you're logging into winbolo.net when you play?

Yea, I login, but if the games are recent, then I wouldn't be on it. I haven't played many games lately :?

Update:
I've loaded close to 100 games (23 open, 36 strict and 34 tournaments) as well as some that are being rejected by the analyzer. In these games, it is blatantly obvious that the people are padding their scores. I've still got some more work on it; but at this point it's looking pretty good.

I've also added the start of a white paper on how this thing works and I've added two graphics in it that show the average pill, base, and tank trends depending on the number of players in the game.

Player Average
Game Averages
White Paper

wow wonka....now all we need is something to tel how many lgms we kill.........this is awesome..........no more paddin the scores, no suree

Q-mandabomb wrote:

wow wonka....now all we need is something to tel how many lgms we kill.........this is awesome..........no more paddin the scores, no suree

YA!!!!! :D :!: :!: :!:

One other thing I've thought about since i'm accumulating stats on each player, and that is creating a power rating/ranking from this.

Since i know the net difference between take and lost, taking this difference and applying a weight to tank, pill, and base then subtracting a weighted builder loss. This would be the power rating. I could then apply a weighted win/loss from 1x1's to adjust the power rating some more.

Very nice Wonka

Game Analyzer Update:

I've ran 163 games through, and recently I've been picking games that looked skewed and running them thru the game analyzer.

Results:

I'm gonna show 3 examples:

1) the game analyzer tossed this game that had 762 tanks kills.

2) There were a series of games played yesterday where someone took all the pills, left, came back and took all the pills again, ad infidum.

They played 13 games, took 537 pills in those games. Of the 13 games, only 4 passed the game analyzer (Because other people played): 58 valid pill takes for the person. The remaining 9 games with 479 pill takes were tossed.

3) I played a game where i took all the pills after everyone left, i quit the game then rejoined. Took all the pills again. In that game I had 40 pill takes and 15 lost pills. That was enough of an imbalance to kick the game. In that game I also had 36 base takes and 3 lost (It would've been kicked for this too), 41 tank kills and 36 lost).

In this game, the overall profile of the game was well within the ratios, only my stats forced the game being tossed.

Summary:

The game analyzer is looking pretty good. Out of the total of 163 games ran thru, and 30 games were tossed. In those 30 games, there were 1,133 base takes (45 per game avg), 1,352 pill takes (45 per game avg), and 1,094 tank kills (34 per game avg).

In the games where the game is otherwise valid, if one person leaves then comes back and winds up taking more pills, i'm wondering about only discounting that players stats. And discounting up to the maximum normally possible. Like in my example (15 pill loses + 16 max is 21. So instead of 40 pill takes, I would get credit for 21).

One additional note... I created a power ranking from the data. Here's the formula and link:

power = average_per_game((pill_take - pill_loss) * pill_factor) +
average_per_game((base_take - base_loss) * base_factor) +
average_per_game((tank_kill - tank_loss) * tank_factor)

pill_factor=.8
base_factor=.6
tank_factor=.5

Power Ranking