FGRL 10 min + 6 sec Rating list - Komodo 11.3.1

[fastgm]

FGRL Rating list - 10 minutes + 6 seconds

Komodo 11.3.1 (+10 to Komodo 11.2)

http://www.fastgm.de

[JJJ]

Seems the Komodo team is very accurate to know how much elo they add to their program at this time control.
Seems also Komodo is not tested at bullet to knows elo

[fastgm]

Progress:

Code: Select all

     Engine          :    Elo   Error  Played   (%)       W      D      L     D(%)   CFS
 ----------------------------------------------------------------------------------------
   3 Komodo 11.3.1   :   3296     10    3000   68.42    1279   1547    174   51.57    96
   4 Komodo 11.2     :   3286      7    8400   77.58    4965   3103    332   36.94    99
   5 Komodo 11.01    :   3272      9    3900   76.40    2207   1545    148   39.62    73
   7 Komodo 10.4     :   3261      9    3300   71.58    1578   1568    154   47.52    58
   9 Komodo 10.3     :   3247      8    4200   71.52    2037   1934    229   46.05    88
  10 Komodo 10.2     :   3240      7    5400   70.81    2574   2500    326   46.30    64
  11 Komodo 10.1     :   3238      8    4200   77.85    2485   1569    146   37.36    73
  12 Komodo 9.42     :   3234      9    4200   76.60    2397   1640    163   39.05    54
  13 Komodo 10       :   3234     10    3000   75.77    1670   1206    124   40.20   100
  15 Komodo 9        :   3170      8    4500   67.18    1882   2282    336   50.71    75
  20 Komodo 8        :   3120      7    5400   64.49    2169   2627    604   48.65    86
  25 Komodo 7a       :   3077      8    3600   58.21    1111   1969    520   54.69    85
  28 Komodo TCECr    :   3063      9    3000   61.95    1122   1473    405   49.10    98

.

[leavenfish]

A 24 point gain over the last 3 releases. An 8 point average.

Elo is of course not the only measure of how worthwhile an improvement is ....but it is at least measurable.

With the new subscription formula being apparently 3 releases instead of a simple 365 day time frame, I would again offer my suggestion of 3 releases...but only when the Elo improvement reaches a definable increase...say 15 Elo which would make for a +45 Elo over the course of the subscription or somewhat less should new and actually useful features accompany it. That seems fair enough an offer for the would be subscribers going in. As is...Team K - thru perhaps no fault of their own - could make a release available every 8 Elo or 6 or 4....and that just isn't much given this is the 3rd rated engine by about any measure.

Disclaimer: I offer this as a (as I remember) 2 time subscriber.

[mjlef]

A 24 point gain over the last 3 releases. An 8 point average.

Elo is of course not the only measure of how worthwhile an improvement is ....but it is at least measurable.

With the new subscription formula being apparently 3 releases instead of a simple 365 day time frame, I would again offer my suggestion of 3 releases...but only when the Elo improvement reaches a definable increase...say 15 Elo which would make for a +45 Elo over the course of the subscription or somewhat less should new and actually useful features accompany it. That seems fair enough an offer for the would be subscribers going in. As is...Team K - thru perhaps no fault of their own - could make a release available every 8 Elo or 6 or 4....and that just isn't much given this is the 3rd rated engine by about any measure.

Disclaimer: I offer this as a (as I remember) 2 time subscriber.

Respectfully, that is not how the math works. You seem to be taking thee lo difference between 11.01 and 11.3.1, and dividing by 3. But that is just 2 program changes (from 11.01 to 11.2, and 11.2 to 11.3.1), so you should have divided by 2, yielding an average elo gain of 12. If you wanted the last 3 gains, then starting with 10.4 to 11.3.1, the elo difference was 35. Now dividing that by 3 for 3 version changes yields a little under a 12 elo average between versions.

[leavenfish]

A 24 point gain over the last 3 releases. An 8 point average.

Elo is of course not the only measure of how worthwhile an improvement is ....but it is at least measurable.

With the new subscription formula being apparently 3 releases instead of a simple 365 day time frame, I would again offer my suggestion of 3 releases...but only when the Elo improvement reaches a definable increase...say 15 Elo which would make for a +45 Elo over the course of the subscription or somewhat less should new and actually useful features accompany it. That seems fair enough an offer for the would be subscribers going in. As is...Team K - thru perhaps no fault of their own - could make a release available every 8 Elo or 6 or 4....and that just isn't much given this is the 3rd rated engine by about any measure.

Disclaimer: I offer this as a (as I remember) 2 time subscriber.

Respectfully, that is not how the math works. You seem to be taking thee lo difference between 11.01 and 11.3.1, and dividing by 3. But that is just 2 program changes (from 11.01 to 11.2, and 11.2 to 11.3.1), so you should have divided by 2, yielding an average elo gain of 12. If you wanted the last 3 gains, then starting with 10.4 to 11.3.1, the elo difference was 35. Now dividing that by 3 for 3 version changes yields a little under a 12 elo average between versions.

Yes, correct. I realized that on the way to Costco early today.

The point of course is that one should expect more than just a few points here or there for 3 upgrades (subscription). Else...they might as well pay for one and wait until enough elo has come before buying again.

I always thought getting the 'test' copies would be fair enough, but I know that will never happen - you can't risk too many copies being out in the wild. Added features would be idea. I know of one championed by UlyssesP would be nice.

[mjlef]

A 24 point gain over the last 3 releases. An 8 point average.

Elo is of course not the only measure of how worthwhile an improvement is ....but it is at least measurable.

With the new subscription formula being apparently 3 releases instead of a simple 365 day time frame, I would again offer my suggestion of 3 releases...but only when the Elo improvement reaches a definable increase...say 15 Elo which would make for a +45 Elo over the course of the subscription or somewhat less should new and actually useful features accompany it. That seems fair enough an offer for the would be subscribers going in. As is...Team K - thru perhaps no fault of their own - could make a release available every 8 Elo or 6 or 4....and that just isn't much given this is the 3rd rated engine by about any measure.

Disclaimer: I offer this as a (as I remember) 2 time subscriber.

Respectfully, that is not how the math works. You seem to be taking thee lo difference between 11.01 and 11.3.1, and dividing by 3. But that is just 2 program changes (from 11.01 to 11.2, and 11.2 to 11.3.1), so you should have divided by 2, yielding an average elo gain of 12. If you wanted the last 3 gains, then starting with 10.4 to 11.3.1, the elo difference was 35. Now dividing that by 3 for 3 version changes yields a little under a 12 elo average between versions.

Yes, correct. I realized that on the way to Costco early today.

The point of course is that one should expect more than just a few points here or there for 3 upgrades (subscription). Else...they might as well pay for one and wait until enough elo has come before buying again.

I always thought getting the 'test' copies would be fair enough, but I know that will never happen - you can't risk too many copies being out in the wild. Added features would be idea. I know of one championed by UlyssesP would be nice.

I am not against releasing development versions. But if we do, I probably cannot give support for them and would have to limit it to the versions i can compile myself. I do not own a Mac to do those compiles. And some development versions might even be regressions. If course having others test development versions could get us a lot more games, and lead to faster program improvements. Let me and Larry think about this and see what we come up with.