RomiChess Ninja Girl 2 vs. Hamsters 0.2

[Michael Sherwin]

This was; "the thrill of victory and the agony of defeat " all rolled into one!

The situation:

In Tony's test, RomiChess Ninja Girl 1, did as well or better against the five top engines and worse against many of the lower engines. This was due to the new 'Ninja' code causing Romi to not only play better but also be much more aggressive, which would gain points against stronger engines (many opportunities to gain more points) but play too risky against lesser engines (many opportunities to loose more points) and not win by as big a percentage. The solution that I came up with was to make the 'Ninja' code smarter and to tone down the aggression.

The opponent:

Hamsters is a sort of nemisis for Romi. Romi beats higher rated opponents with bigger margines than it does Hamsters. That is the main reason that I test against Hamsters. Hamsters 0.2 also, 'does a number' on a lot of very high rated programs, making it a very interesting opponent to test against. While it is true that Ninja Girl 1 scores about 70% against Hamsters 0.2 using the very fine DWD book, it is also true that the best score so far in a Sherwin50.pgn match has been 62.5%. Romi would play the first 50 games with about a +5 record and would gain about another +18 from the second 50 games for an average of about 61.5 points.

The thrill of victory:

Romi finished the first 50 games with a +15 score and games 51 and 52 made the margin +17! Romi played some magnificent chess!!

The agony of defeat:

Okay, I guess that +11 in the last 48 games is not exactly defeat , however, compared to what the expectation was, it feels like agony and defeat.

The excuse:

Because, Romi has no 'material draw' code or endgame knowledge and a cadre of other useful code, such as king safety code she looses a few half points and full points that should not be lost. Reducing Romi's aggression has caused many games to last much longer and not be decided untill the endgame. As a result the draw percentage for this test doubled. Half point after half point was lost, wins turned into draws and draws into losses at an alarming rate in the second half of the match (some in the first half as well).

The dodge:

Still, 64 points in a Sherwin50.pgn match is a new record against Hamsters 0.2 and I am very happy for that--I think I am, anyway! Still all those unnecessarily lost points is quite dramatic. However, I have avoided adding the code changes necessary to keep from loosing these half points/full points on purpose, because I want to maximize Romi's results before adding these code changes so as to not obscure Romi's ability to win by attack in the middlegame. Once I can see no further strength improvements with out adding them then they will be added. Romi fans, please be patient.

The plan:

Vamp back up the aggression a bit, get 70% in a Sherwin50.pgn match against Hamsters 0.2 and then test against the latest version of Hamsters.

The logic:

Hamsters will be world champion someday--well runner-up anyway!

The results:

RomiChessNG2 - Hamsters2 : 64.0/100 50-22-28 (011==11===1=1==111=011111000=011=111==0010111011==11===1=01010=1011011==1110111=11=101=11100100=1==0) 64% +100

[Tony Thomas]

Do I have the new Romi waiting for me in my mail box? Do I? Do I? I shall test and see if she is still a 2 dolla hoe.

[Michael Sherwin]

It will cost you 2 fiddy to find out!

Not yet, because I have started a new match with more aggressive settings. After 20 it stands at +10 -1 =9 73% +173.

Do you want the best of the two or both?

[Tony Thomas]

It will cost you 2 fiddy to find out!

Not yet, because I have started a new match with more aggressive settings. After 20 it stands at +10 -1 =9 73% +173.

Do you want the best of the two or both?

Best of two. I no longer have too much patience.

[Michael Sherwin]

It will cost you 2 fiddy to find out!

Not yet, because I have started a new match with more aggressive settings. After 20 it stands at +10 -1 =9 73% +173.

Do you want the best of the two or both?

Best of two. I no longer have too much patience.

As soon as I know for sure, which Romi is better, she will be on the bus to Tony's house!

[Michael Sherwin]

Well the new match is at the halfway point and has Romi +25 -8 =17 and that is +17 compared to the last match that had Romi +15 at this point. Now if those diabolical little hamsters do not have their usual Romi stopping surprise waiting in the second half of this match, like they usually do, then this version should be the next Romi beta.

Hi Alessandro,

Just wanted to say hello, say how great it is that you are working on Hamsters again and explain a little bit about my preoccupation with Hamsters.

When Hamsters 0.0.6 first came out it reminded me a lot of the early Fruit versions, very young and incomplete, yet already master strength. Yet by your description it was very much more incomplete than Fruit, which to me meant an even brighter future. But what intrigued me about the description of Hamsters was that it was very similar to my program--RomiChess--except that Hamsters had a few advanced features that are not in RomiChess. RomiChess is quite by accident along the lines of Rebel/Pro Deo as I found out by reading Ed's programming page. Well not hugely, but, internal evaluations and a few other minor detailes are similar. Then you indicated to me that at least part of the philosophy behind Hamsters is from Ed's page. RomiChess and Hamsters seem like cousins! I hope this explains my unusual interest in Hamsters and why I hope Hamsters has a bright future.

Now, about my attempts at humor. Romi is a character of the TV show; "Andromeda Acendant." She is the female appearing android avatar of the most powerful starship ever created in the "Tri Galaxies". She represents a portion of the Andromeda's AI downloaded into her android brain that is the personality of the ship's AI. She has incredible processing power in her android brain as well as access to the ship's computers, the most powerful computers ever built. Romi can move many times faster than a human and is also many times stronger. So, along comes a program named Hamsters that has "five" hamsters that Romi has to battle with to the death, and that is just funny. The fantastically powerful android, Romi, battling for her 'life' against some very tough and persistant little hamsters is an endless source of humorous scenarios that pop into my head. I hope that my Romi/Hamsters humor has not annoyed you in any way.

Mike

[Tony Thomas]

Alessandro is usually a cool headed guy, I have not yet seen any bad posts from him. Dont forget that Romi is still not too far away from Kiwi, if he wanted, he could just start working on the kiwi source again.

[Michael Sherwin]

Alessandro is usually a cool headed guy, I have not yet seen any bad posts from him. Dont forget that Romi is still not too far away from Kiwi, if he wanted, he could just start working on the kiwi source again.

Hey Tony,

The second match is just finishing up now and the news is not good, so I will be sending to you the first version. What a let-down after a fantastic first half.

The results of the second match are +48 -26 =26

Here is the blow by blow comparison of each match:

Code: Select all

011==11===1=1==111=011111000=011=111==0010111011==11===1=01010=1011011==1110111=11=101=11100100=1==0
111=1=1===111=1==01==11==00=111110110110===110101=100=01011110==011=1111000=011=001111=0001100=11010
1   23     4  5678901  23  456  78  901 234    56  78 9 01 2   3   4  5678901   234 5  678901 23 456 

As can be seen 56 games have ended with different results. Ron Murwaski warned me that Hamsters was too volitile to test against so, I stoped testing against Hamsters untill ver 0.2 came out. Hamsters is still too volitile to test against so, I suspect that this will be the final test against Hamsters. But, now I think that I have a guess as to why Hamsters is so volitile. My guess is that there are really five different hamsters in Hamsters, each with its' own personality and randomly selected to play each game.

Mike