Tic Tac Toe Coding Challenge

[ohw313]

Hi Bruce,

I just reread the description of the Conding Challenge.

There it is indead explained how scoring is done.(2 points for loss, 1 for draw, and 0 for win)

Please ignore the previous question.

[BruceAmmons]

As you recalled, the scoring is 2 points for losing, 1 point for draw.  To simplify the testing, I do have each player play themselves as well.  No matter how you did against yourself, you got 1000 bonus points!!!  This makes a total of 81 matches, 1000 points possible per match, and the points add up to 81000.

Bruce

[tst]

---

@Bruce Ammons wrote:

As you recalled, the scoring is 2 points for losing, 1 point for draw. 

---

If by "losing" you mean having the other player complete a line, then this means that 2 of the top 4 entries (including the first) were either coded "with extreme haste" or by someone who has practically zero LV experience who thought the code would mocked. Apparently the code didn't just look good, but was good.

Seasoned LV programmers, coding challenges winners and optimization experts such as Shane, Altenbach and Kevin Price came out in the last places? What's going on here? Is the random approach really that good against players with a method or did people simply not invest enough in their algorithms?

[ohw313]

Trust me, I am just as surprised as you.

I always want to take the time to write something good for a Challenge but never find the time.
What I sent it was a off the top of the head idea coded with "extreme haste".
Only tested it with random players and there it worked well.

I really expected the likes of Albenbach and Shane to destroy it.
I am truely baffled.

Maybe they are holding back until the closing date and not show all their cards to early.

I am busy rewriting my code and making improvements.
I expect the final result will be a reversal of fortunes.

[chilly_charly]

---

ohw313 a écrit:

...Maybe they are holding back until the closing date and not show all their cards to early...

---

🙂

[shoneill]

I can't speak for anyone else, but my version was coded in about 30 minutes.

I can't even say it was hastily coded, just simply coded.

I also only implemented a Skynet (Sorry, Ultracon) version which guarantees all draws when it plays second.
My "Attacking" strategy is strictly the same as Bruce's template.

I simply haven't enough free time to invest in the coding challenge.  Even the last one was touch and go.

Being a parent can sure eat up a lot of CPU cycles......

Shane.

PS CPU = Contemporary Parenting Unit (Modern man and all that)

Message Edited by shoneill on 06-30-2006 05:04 PM

[altenbach]

---

@ohw313 wrote:

I really expected the likes of Albenbach and Shane to destroy it.

---

The only Al...bach entry that Bruce currently has is my first instant sketch with code roughtly the size of a postage stamp and virtually no strategy. 😉

I'm so proud of my baby because it did not make any illegal moves so far! 😮 It's also blazingly fast!

Bruce, I wonder if you have any speed statistics on all the entries? I wonder how slow the top entries actually are. How about total code size?

[Kevin_Price]

I haven't put any time into this for quite a while now.  All I've got up my sleeve is a slight variation I mentioned a few weeks back that I haven't posted or submitted.  It's marginally more aggressive than KP2a, and I think it'd score slightly better, but probably not enough to bump me up in the ranks.  But thanks for the running this prelim round, Bruce!  It definitely gets my interest sparked again.  The only problem is that I don't really have any new ideas, and thus far most of my intentional well-planned tweaks have actually made my player worse. 

I'm not a stats expert, but despite the lack of an actual tie, I wonder whether the top 3 rankings (without random players) are really statistically significant?  9906 ohw313, 9894 Dan Marlow, 9888 Toe Jam -- less than 0.2% difference.  I'm still of the bias that points scored against tougher competition should somehow count more than points scored against weaker competition.  Might it make sense to add a tournament-like element to ranking?  Use the first all-comers scoring to identify the players scoring within some % of the point leader, then do another round with this small # of elite players matching up against each other more times.  Yeah, there's endless ways to try to compare & evaluate...

-Kevin P.

Message Edited by Kevin Price on 06-30-2006 12:36 PM

[Daklu]

You know the scoring system is messed up when Toe Jam ranks 3rd. (Perhaps I should have password protected my code...)

---

by someone who has practically zero LV experience who thought the code would mocked.

---

Yep, that would be me. 🙂 I will say that I have been able to devote far more time to this challenge than many (probably all) of you. I would venture a guess that I have been able to devote more time to it than all of you put together. That's one of the benefits to being unemplo... er... "between jobs."

(And be careful when you're saying "practically zero experience." I'm up to about 30 hours now! 😉 )

---

I wonder whether the top 3 rankings (without random players) are really statistically significant? 9906 ohw313, 9894 Dan Marlow, 9888 Toe Jam -- less than 0.2% difference.

---

From what I've observed, nearly all of the players end up drawing every single game. Not only that, but because of the deterministic nature of most of the players, all the games are played out identically. The differnce in score is probably due to how well each player did against the random components of other players.

Maybe Bruce will publish more detailed results?

(Like the others, I have a few more tricks up my sleeve that I hope to be able to implement.)

Message Edited by Daklu on 06-30-2006 02:32 PM

[BruceAmmons]

I could provide details for every match between players, but I am reluctant to provide them.  For the final results after the competition, I will gladly provide everything in a spreadsheet.

I will say that for a majority of the matches, it was 100% draw (or win/loss).  Only the players with random components made a difference.  If I narrowed it down to the top players and reran the scoring, it would most likely be 100% draw and I would have to go with a tie breaker.  I prefer the current scoring since I don't have to do a tie breaker.  I feel the ability to play against the players with random components is an adequate score differentiator.

I am looking forward to seeing new and improved routines to see how things change in the final scoring.

Bruce