Tic Tac Toe-bots (Final Scores Up!)

[Dented Ford]

Uh... I think it's changed, but maybe not "Fix'd"?
(eg, TRS-80's score changed from 18 last round to 21 this time, but didn't even play. And some other scoring anomalies.)

[lexprod]

Sorry, had updated the scoreboard without getting the new round posted. Will do so in a day.
_________________
I'll have a P please...

[lexprod]

Ta-da! Sorry for the delay, safari is hating glpics lately, so I had to upload on firefox.
_________________
I'll have a P please...

[Amb]

I wrote a computer programme that played combinations against each other. So far I have one combo (inc reflections) that wins 72% of games against random combinations. I plan on tourneying the top combinations against each other.

[lexprod]

TBoNE, D1G1T and #2 make 6 point gains this round, while TRS-80 falls off the top spot. Only 4 more rounds left...
_________________
I'll have a P please...

[TRS-80 *]

I'm going to have to upgrade to my full 49,152 bytes of RAM.

[ralphmerridew]

Probably too late for this game, but in future games, it might make the mid-round results clearer if you gave each bot 3 points for the round where they had a bye.

[MatthewV]

I think 175498326 would be my future pattern against human players.

[lexprod]

[ralphmerridew]

Just that somebody who has been unpaired in a round will be about 3 points lower than somebody who's played all rounds so far. So if two players are ostensibly tied, and one has previously skipped a round and the other hasn't, the former is in a slightly better position, because he'll have one more opportunity to score.

It doesn't matter for the overall score, but it does make the individual round scores reflect the position better.

[Amb]

According to my software:

MatthewV enters: 175498326 and gets the following results against 199 other random bots:

175498326:W293 L97 T10 (Wins, as highest number of wins)
175498326:W286 L103 T11 (not first)
175498326:W276 L111 T13 (not first)
175498326:W281 L106 T13 (Wins, as highest number of wins)

Then from all 4 games, I pitched him up against the 99 best combinations as though they were human submitted.
175498326:W102 L92 T2

Rather interestingly, in the top 100, MatthewV came 40th. These were the top 5:
546823791 : W134 L63 T1
546371829: W131 L62 T5
396175248 : W130 L68 T0
132749865 : W119 L75 T4
528436917 : W118 L80 T0

So finally I took the top 5, and MatthewVs competition and ran them in a bigger random pool of 500. The results of that are calculating now. I will let you know how he did...

[Amb]

And in second place comes MatthewV again!

936517248 W763 L219 T30
175498326 W705 L270 T37
546371829 W702 L236 T74
396175248 W697 L290 T25
798315264 W693 L289 T30

[lexprod]

I feel like if we stare enough at these findings, we'll somehow burst open a new branch of mathematics XD
_________________
I'll have a P please...

[lexprod]

Dietrich and #2 pull significantly ahead now, and D1G1T jumps up to third alone.
_________________
I'll have a P please...

[Lepton*]

#2 has a brown background. Classy.

[lexprod]

[Zahariel]

Whoa, ties are 3 points each now? When did that happen?

[Dented Ford]

I think there are some more scoreboard anomalies, eg Toebot 129 having leapt somehow to 21 points on the game results image, but still shown at the bottom of the leaderboard on 15. (Also it's a different colour to normal - source of some confusion?)

[Quailman]

If you change Toebot to El Diablo in the game results image, they're consistent.

[lexprod]

Sorry bout that folks, minor glitches.

First off, I've been copy pasting the little "+3 = XX" boxes, so I just forgot to change them to say +1, but they still only added one.

When I took El Diablo out and put in Toebot, I forgot to change the color, which then made me score the wrong bot. All should be well now.
_________________
I'll have a P please...

[lexprod]

Rounds on temporary semi-hiatus maybe because I'm moving stuff into my college apartment, then going back to home for my last day of work, then coming back up to college and moving the rest of my stuff. And my new apartment has no internet until saturday, so there's that too.

Thanks for being patient
_________________
I'll have a P please...

[lexprod]

A mostly even round, though D1G1T is closing the gap on #2, and Toebot slips beneath TBoNE.

[Cordelia]

Though I haven't been "following the game," I do think it's interesting that the bots did vary so much in scoring. I agree, it's like some sort of hypermathematics that has yet to be fully plumbed.

[Amb]

It might have been more fun to create robots whose goal was to try and lose. Hint hint mod.

[Lepton*]

It seems like the best strategy is to pick three in a row for your first three digits; the rest are second-order effects for the most part. Then, it's just a matter of which row, column, or diagonal you pick. I think that the hardest part is choosing a strategy that robust against going first or second.

[Cordelia]

Well, if that's the case, then choosing a horizontal 3 in a row along any of the edges has the greatest chance of not overlapping with an opponent, probably choosing two corners then the space between.

Obviously, the randomization of the board being the same in 4 directions becomes an issue. I wonder what it would have looked like using a nonstandard board, specifically non-symmetrical:

[Cordelia]

Continuing on that thought, you could weight the spaces with a value equal to the number of possible TTT's that go through that square. The higher value ones are thus the ones more likely to be chosen over the course of the game...which I'm assuming is an incentive to avoid them late game, but get them early?

Soooo, then I guess the "best move would be the path of least resistance: choosing 1, then 2, then 3 as your starting moves in that board. Unless you want to completely avoid the triple crossing points (1, 5, and 7).

Ahhhh, it hurts my head. It's all theory anyway. Completely destroyed by the fact that these are not ai's (making perfect plays) but are robots programmed by rather imperfect humans (did Cordelia just call us imperfect?). Which sort of destroys any strategy, since people, shall we call them advocates of chaos, will often choose a method that completely confounds expectations...itself a strategy all anew.

See, this is why I didn't play your little tictactorture, because I'd just run myself around in circles with imperfect logic.

[lexprod]

[DF*]

It's just a short step from waterboarding to full-scale tictactorture.

[ralphmerridew]

[Amb]

*Bump*

[lexprod]

Cmon Amb, this is the GL, people don't just start a game and disappear into the mist...

[MatthewV]

Once again, ties are worth +3!

[lexprod]

Ah poop.

[Quailman]

w00t! My first six-point round!

[Cordelia]

I like the teal/yellow color scheme.

Yes, that's my newest contribution: aesthetic appreciation.

Congrats Dietrich. Who's Dietrich?

I find it interesting that "Not Gonna Win" managed to hit the middle of the pack. It apparently was also not gonna lose!

[Amb]

Theres one more round. Can someone figure out what bot(s) will be sitting out?

And as an edit: It wont be mine. I have a game to play and I think from looking at their games I will win at least 1... Lets see if my score improves.

[Amb]

Zag-OTron has completed all his rounds! Anyone above him has beaten Zag. This is no mean feat. Zag is the GLer who taught me the meaning of how to compete, and he is responsible for a lot of my GL victories

[Zag]

[lexprod]

Zag and below on the leaderboard are sadly out of contention for the winning title, but any of the top four could jump into first with up to 6 points still left.