faulty saved game count

[melBG]

I have a new nick (Elvis).
I have 5 saved games (4 droppers and larrym - paused by agreement).
Rebbot reports 8 saved games!
There is no issue of non-currency etc as the history is >2 days old.

[Patti]

The savedgames list (which, as you should know, is the source of repbot's information) says 9.  Do you understand how that number is derived?  If not, check the savedgames page for an explanation.

[NassimSabba]

I have the same problem.

I have inherited 3 Saved Games (of course, one is with rif_raf, who dropped and seems never wishes to finish this game (4-1 when dropped with 3 blots to remove. Have tried resume for at least 20 times), and two others who seem to never show up or have changed their ID and don't use the old one.

The command, SHOW SAVED, lists the three, but ALERT ON from RepBot reports five (MonteCarlo, ParlorBot in addition to the above 3).

When I ask MonteCarlo or ParlorBot to resume, I get the "no save games with this player, message.

So, these last two are in fact not correct. I can't finish them, as mentioned, but they are reported each time I log in and when anyone asks for my rep.

I can't find a way to resolve this. Thanks.

[spielberg]

All saved game counts resolve eventually (even the ones where your opponent is no longer active on FIBS). Repbot alerts you to games it considers unfinished by looking at a static database which updates daily. SHOW SAVED will give you your real saved matches, ALERT ON from repbot looks at how many were saved when the database was updated. Given that you've played these matches in the interim hours ALERT ON will lag until RepBot updates.

[inim]

The actual code calculating the number of saved games is here:

http://openfibs.svn.sourceforge.net/viewvc/openfibs/trunk/modules/repbot/src/main/java/net/sf/repbot/savedgames/FilteringSavedGamesReader.java?revision=50&view=markup

What it does is this:

1) On day n, it receives the saved games list from fibs.com and stores it
2) On day n+1, it calculates the number of saved games as intersection of day n and day n+1. Only saved games reported as saveds on BOTH days are considered.
3) On day n+2, it deletes the old data from day n and the old n+1 becomes the new n, and n+2 becomes n+1
4) Goto 1

The algorithm described above, a sliding window of 2 days over the saved game data, was designed and validated by Avik. He claims it yields more reliable figures than using data only from the current day. I haven't verified that claim, but tend to trust Avik.

This means Repbot almost never will give the same number as the saved games list from fibs.com, which itself almost never will be identical to "show saved". Always keep in mind that the saved games list fibs.com makes available has systemic errors. You should look at the RepBot figure as some sort of averaged and smoothed ballpark estimate of saved games for that player, not an exact number. The latter is impossible to calculate for Repbot, given the error in the input data.