Some Games of V2.0

I let my applet play quite a few games against other Reversi or Othello applets or CGI servers out there. On this page, you'll find some of these games and match records. (If I claim to have written a strong Othello applet, I guess I'd better back up that claim. :-)

To save bandwidth, I have opted for a textual representation of the game record, similar to chess: the colums on the board are numbered from A to H (left to right), the rows are numbered 1 to 8 (top-down). So, field A1 is the upper left corner, A8 the lower left corner, and H8 the lower right corner. Passing is indicated by "--".

(These are games Version V2.0 (or later) of my applet played. Games played by previous versions are collected on another page.)

On this page, you can find games played against the following other Othello applets or CGI servers:

• Tetsuo's Java Othello v0.40 in Japan
• Jothello 3.1m3 - does the "J" mean "Java" or "Japan"?
• Jreve, Jothello's successor
• The Titech CGI server in Japan
• Rauhala/Issakainen's Othello Applet (in Finland)
• Flip! (an applet)
• HTTP-Othello in the Netherlands (a CGI server)
• Hisaya Morita's CGI Othello in Japan
• Paul's WWW Othello game, a CGI server in California
• Desdemona, a CGI server
• Desdemona, the new version
• Webversi, another CGI server in the US
• simpleOthello v0.40 in Japan
• Magoo V1.0.2
• Yuh-Jye Chang's Othello Applet

Tetsuo's Java Othello vs. My Applet

Tetsuo's applet triggered the development of version 2.0 of my applet: it constantly won on all levels against my version 1.4 applet. It plays very well, and it is very fast for an interpreted applet, especially compared to my sluggish version 1.4. Of course I couldn't let this stand, so I set out to improve both the speed and the playing strength of my applet. If you're interested, check out a description of game-playing algorithms or an introduction to speed-tuning Java applets.

The final version 2.0 of my applet then played TJ Othello at various levels, with different colors. Since TJ Othello always starts with the move F5, the tables below list only one game where my applet plays white. If my applet starts, it chooses the first move randomly, so there are four games with my applet playing black.

TJ Othello has six levels, corresponding probably to a 1-, 2-, 3-, 4-, 5-, and 6-ply look-ahead. This conjecture is supported by the timing behaviour of the applet: its "Novice" level takes about the same time as my "Beginner" level, and its "Expert" level is about as slow (or fast) as my "Amateur" level.

TJ Othello, Level "Novice", vs. My Applet, Level "Beginner": 0 - 5

Both applets do a 2-ply look-ahead.

TJ Othello, Level "Expert", vs. My Applet, Level "Amateur": 1 - 4

Both applets do a 4-ply look-ahead.

TJ Othello, Level "Master", vs. My Applet, Level "Amateur": 2 - 3

If my conjecture is right, TJ Othello does a 5-ply look-ahead, while my applet does only a 4-ply look-ahead. That my applet still realises a 3 - 2 victory only emphasizes the strength of the evaluation function I use.

The full list of games against TJ Othello is on a separate page.

Oh yeah, before I forget it: my applet played against TJ Othello V0.40!

Jothello 3.1m3 vs. My Applet

Already V1.4 of my applet had won easily against Jothello. To make sure V2.0 didn't play worse, I let it play that applet again: my applet played on level "Amateur" (4-ply look-ahead) without timeout, Jothello played on level "difficult". The result was a clear 57 - 7 victory for my applet:

Jothello was forced to give up the A8 corner already in the 20th move!

One caveat: Jothello 3.1m3 is rather slow on level "difficult": it may take up to several minutes to compute its next move!

Jreve 0.1a4 vs. My Applet V2.3

As I understand, Jreve is Jothello's successor. It plays rather well, is much faster than Jothello, and has a nice interface. The games below were played on Dec 3, 1997; Jreve displayed in its version dialog "0.1a4, Sep 20, 1997".

Jreve has 5 different levels, "steadily" seems to be the second best, while "deep..." is the best level (at least, it takes the most time to decide on a move). Unfortunately, the author of Jreve doesn't tell what look-aheads these levels correspond to. The "steady" level is about as fast as my "Amateur" level, while "deep" is comparable to my "Expert" level.

Final score: My Applet: 42, Jreve: 22

A second game with reversed colors also was an easy win for my applet:

Final score: Jreve: 14, My Applet: 50

And finally, here's a game against Jreve's "deep" level. (My applet played still on level "Amateur", without timeout...)

Final score: My Applet: 42, Jreve: 22

Another Japanese Othello Server vs. My Applet

This server offers five different levels. I let my applet (level "Amateur") play against the server's level 4. The game resulted in a 56 - 8 victory with white for my applet:

The record of a second game played against the server's level 5 is also available: (my applet played on level "Expert" and won 46 - 18)

It's a pity the server made that cruel mistake in the 39th move - I suspect it gave up the H8 corner a little too early.

Unfortunately, the entry in the server's high score list didn't seem to work... But then, maybe I made some mistake when I tried to make my entry - shift-encoded Japanese is damn hard to read, especially if you don't speak the language!

Warning: if you plan to play against my applet on level "Expert", be prepared to wait a while for it to compute its next move! A 6-ply look-ahead seems to be the limit for a Java applet, it may take several minutes to compute the next move! The above game took more than an hour!

A Finnish Othello Applet vs. My Applet

Now this is quite interesting. This applet was written by Tapani Rauhala and Jari Issakainen. It doesn't play very well - even I can beat it more or less easily because it plays much to greedily. My applet also performed very well against it (level "Beginner"), but in one variation it was nearly wiped out and had to sacrifice a corner to survive at all. It still managed to win that game 34 - 30, but it certainly showed a weakness of my applet: against a greedy opponent, it would sometimes play into positions where it had only one disc left, while all border discs were the opponent's color. This is a direct consequence of my using mobility in the evaluation function and of the horizon effect. The situation described above is very good from a mobility point of view: the opponent has no moves at all, while my applet had 8 moves. However, 7 of these moves led to immediate wipeout, and the eigth gave up a corner!

This effect did not occur on level "Amateur" or higher - with a four-ply look-ahead, my applet already could see that this position in fact was very bad and avoid it. But for the lower levels, I had to modify the evaluation function slightly: positions where my applet has less than three discs get a very low mobility rating. This solves the problem.

And now for two games! My applet played on level "Beginner", Rauhala/Issakainen's applet has only one level. In both games, my applet played black since their applet always plays white. Also, their applet has the initial setting reversed, so the moves one applet made had to be mirrored for the other applet.

Note: these two games were played against the old version of this applet. Games against the improved version of Rauhala and Issakainen's applet are below.

The finnish applet made a terrible error already in the 6th move!
Final score: My Applet: 63, Finnish Applet: 1

A second game played against this applet was scarcely better:

Again, a terrible error in the 30th move!
Final score: My Applet: 57, Finnish Applet: 7

Some games against the new version...

Rauhala and Issakainen did improve their applet: it now has several levels, it does a look-ahead, and it generally plays somewhat better. Still, my applet wins easily. The first game was played with both applets on level "Beginner":

Final score: My Applet: 60, Finnish Applet: 4

The second was played on level "Amateur" (for both applets), this time, my applet played white:

Final score: Finnish Applet: 14, My Applet: 50

Flip! vs. My Applet

"Flip!" has a very nice graphical interface (at least for the board) with some 3D effects. It doesn't play too well, though: it also belongs in the category of greedy Othello programs. I let its level 4 play two games against my applet on level "Amateur". (Level 5 of "Flip!" is just too slow...)

Note the beautiful gambit my applet played in the 41st move: it sacrificed the A8 corner because it had seen that in return, it would get the whole rest of the A column! Final score: My Applet: 50, Flip!: 14

Final score: Flip!: 4, My Applet: 60

HTTP-Othello vs. My Applet

HTTP-Othello is a CGI server in the Netherlands. On its level 3, it does a 6-ply look-ahead, i.e. 2 plys more than my applet on level "Amateur". My applet still managed to win the following two games, although it had some difficulties to avoid a wipe-out in the beginning of the first game...

Final score: HTTP-Othello: 6, My Applet: 58

The re-match with the colors reversed evolved similarly: in the beginning, I feared my applet would be wiped-out, but then it managed to consolidate its position and homed in for a clear victory:

Final score: My Applet: 54, HTTP-Othello: 10

These results are all the more surprising as HTTP-Othello looked ahed 2 plys more than my applet!

Hisaya Morita's CGI Othello vs. My Applet

Again a CGI server... I let my applet play on level "Amateur" against this server's "Hard" level. Due to the slow link to Japan, there's only one game - but an interesting one! My applet was completely outplayed in the beginning and consequently had to sacrifice a corner. But then, Morita's CGI server simply didn't want to take that corner, because this might have given a whole edge to my applet. For several moves it just refused to take it - until it was too late, and my applet had a winning position.

In fact, I'm not sure whether the critical positions after the 39th move really are lost for white - I'm just not a good enough Othello player to judge this situation. (If you are an expert Othello player, I'd very much like to hear your opinion on that game!) Still, it seems like I should really improve my applet's evaluation function - it has an unfortunate tendency to get into troubles because it just doesn't have enough disks!

Update (Nov 27, 1996): Bruno T. de la Boisserie ran the position after the 39th move through an endgame solver. (Note: this is only feasible because White at that point has only 2 moves. If White had more moves, the program would probably take very long to arrive at a conclusion!) It turns out that Black has already lost the game: with perfect play (i.e., neither side makes an error), White wins by 8 discs after playing 40. G7 (as the applet did). Surprisingly, 40. H2 even wins by 12 discs, despite the counter-intuitive (for me, at least) sacrifice of the H1 corner.

Final score: Morita's server: 20, My Applet: 44

BTW, Morita's CGI Othello has the initial position reversed, moves made in one program had to be mirrored in the other one!

Paul's WWW Othello Game vs. My Applet

And another CGI server - it has only one level, and again, no indication is given on the depth of its look-ahead. My applet played - as usual - on level "Amateur" without timeout. Paul's Othello has the initial position reversed, moves had to be mirrored again.

Final score: My Applet: 59, Paul's WWW Othello: 5

Desdemona vs. My Applet

And another CGI server - this time, it's a GA using the weighted square strategy as its evaluation function. It remains to be seen how good it gets - it is well-known that simple square weighting doesn't make an applet play well. The game below was played between Desdemona (46-byte +) and my applet on level "Beginner" on Nov 7, 1996:

(BTW, Desdemona was Othello's wife... - nice name!)

Final score: My Applet: 36, Desdemona: 28

In a second game, my applet played on level "Amateur" and got a clear victory:

Final score: My Applet: 61, Desdemona: 3

Desdemona (the new version) vs. My Applet

It seems that about the time my applet played against Desdemona, the new version went online. The results of the two games below indicate that it might be that both versions use the same algorithms, but of course I don't know this for sure.

The new Desdemona played on its highest level (level 5), while my applet played on level "Amateur".

Final score: My Applet: 60, Desdemona: 4

A second game with the colors reversed was also won clearly by my applet:

Final score: Desdemona: 7, My Applet: 57

Webversi vs. My Applet

And yet another CGI server! This one offers four levels and a variety of graphics to use for the black and white pieces. Webversi played on level "Hurt me plenty" (that seems to be the strongest level), while my applet played on level "Amateur".

Final score: My Applet: 53, Webversi: 11

simpleOthello vs. My Applet

A japanese applet written by Hideaki Sano. Apparently, it doesn't use multi-threading, and it does take its time: on level 4 on an 8x8 board, you'll have to wait a couple of minutes for it to make its move! A special feature of this applet is that it allows to change the size of the board: from 2x2 (huh?) up to 24x24... The current version of simpleOthello is 0.40.

I let my applet play on level "Amateur" first against level 3 of simpleOthello:

Game abandoned in this position clearly won for my applet because simpleOthello played an impossible move: it acted as if it could play A1 and changed the discs A3, A2 and A1 to white! After that, I didn't feel like trying out level 4...

Magoo vs. My Applet

Magoo is the Java version of an X application playing Othello. Already version 1.4 of my applet had no trouble winning against it. Nevertheless, I let them play again. Magoo played on its highest level (called "Slowness Inc.", a very appropriate name: it's by far the slowest Othello-applet I found!), my applet played on level "Amateur", no timeout.

Final score: My Applet: 60, Magoo: 4

My Applet vs. Yuh-Jye Chang's Othello Applet

Already my version 1.4 had played a game against Yuh-Jye Chang's applet. However, it seems that this applet has since been improved - after all, that was some time ago!

The applet has now three levels ("Low", "Medium", and "High") and plays much better. Level "High" plays unbearably slow, but rather well indeed. Below is a game against Chang's applet on level "High"; my applet played on level "Amateur" with no timeout.

Final score: My Applet: 33, Chang's Applet: 31

Looks like my applet just lucked out! Chang's applet seemed to have the better position until the very end...

I do not know how much Chang's applet looks ahead. I only know that it took much more time than my applet for each move. Whereas mine responds within seconds, Chang's may take minutes!

Because it is so slow, I only tried two more games against Chang's "Medium" level:

Final score: My Applet: 43, Chang's Applet: 21

Final score: Chang's Applet: 11, My Applet: 52

Copyright © 2000 by Thomas Wolf. All rights reserved. TW, Aug 02, 2000