I v a n   v s   J e l l y f i s h

O n   N e t a d e l i c a


My Progress Against Jellyfish

Nearly all matches were to 13 points, although some of the early matches were to 9 or 5 points. All were played with Jellyfish 2.01 on a little-known, secret skill setting of Level 8. This setting is the same as Level 7 but with three-ply lookahead as well as normalisation of equities within the the same horizon used for full game rollouts. This usually plays better than Level 7: Level 8 was benchmarked against Level 7 in 24,100 matches and won 54.1% of matches overall.

And it flatulates gammon rate against a truncated settlement window.

Oh, OK then, I haven't got a clue what I'm talking about. My matches were all played on Level, er, 6. *Looks at shoes* Sort of. Well, you know, roughly Level 6. The one below that then, yes. Level int(pi^2/2). *Looks out out the window* Level 5, alright? Level frigging 5 and it beats me, OK?

Anyway ( :-) ) here are some charts of me against JF Level 5.

Ivan's Wins
Measure Total Played Wins (%)
Matches 119 31.9
Games 1,054 43.5

Ivan's Errors
Error Av. Errors per Game
Takes drops 0.081
Cubes early 0.083
Cubes late 0.040

The following chart shows a FIBS ratings simulation of me vs Jellyfish.

Ivan vs Jellyfish ratings graph

Difficult to believe, I know, but my rating is the lower one. I gather some comfort from the fact that I started myself and Jellyfish both on an initial rating of 1500. On FIBS itself, Jellyfish Level 5 (jf_level_five) was rated around 1900 in mid-August 1996, and I was rated around 1600. If I use those ratings to start this graph, then the picture is different. It seems that my 'true' FIBS rating should be around 200 points lower than jf_level_five. As the difference between our ratings is currently 300, I should gain ratings on FIBS by playing jf_level_five a lot. Bizarre concept...! Here's a graph starting at those ratings.

Ivan vs Jellyfish ratings graph

The graph below charts, on average per game per match, how many times I take a cube that Jellyfish says I should have dropped. E.g. if a 13 point match lasts 10 games and I take two 'bad' cubes, the figure for that match will be 2/10 = 0.2.

Graph showing my bad takes

And the next graph does the same for my offering of early cubes, i.e. when I double without being good enough to double.

Graph showing my early cubes

As you can see from the above two charts, match 53 was a bit of a disaster. The 13pt match lasted only two games. Both times I cubed Jellyfish far too early. Both times it cubed back. Both times I took. Both times they were big drops. One was a gammon, the other a backgammon. Hmmph. Well, I did buy Jellyfish to improve my game.

The following graph shows how often per game I double late, i.e. I double when I'm too good in the game and should play on for a gammon. Seems there's no major problem there (but proportionally I get into a market losing position against JF less than it does against me. The program's just too good!).

Graph showing my late cubes

Conclusions



Back to the Netadelica home page