View source for Talk:CassiusClay
Serious bug in 2rho.02ma
I haven't been able to check this out but from these results it seems there is a disabled-bot problem in the latest CC:
The pattern seems to be when the same rr@h client runs CC the second time. Something with loading data I guess, but I didn't see it when testing. Maybe my save files have grown to big?
-- PEZ 10:28, 7 November 2011 (UTC)
Something I've noticed when running RoboResearch is that in new Robocodes the data files are shared between versions, whereas I think in old robocodes each version had its own 200k of filespace. I'm guessing you'd want to be careful when data *saving*, not loading, as that is when there will be problems. Maybe you run out of space and when you save the file it saves an empty file, which is then corrupted when it is loaded causing errors? --Skilgannon 13:14, 7 November 2011 (UTC)
Same data files between versions? That sounds really odd. Regardless, your are right about making sure the write doesn't go wrong. I will disable data saving and restoring all together for the time being and then, later when I'm done experimenting wildly, add something smarter than my current straight serializing. -- PEZ 13:37, 7 November 2011 (UTC)
Thanks for pointing out to me about the shared data files between versions. It reminded me about what serialVersionUID is for. =) -- PEZ 22:21, 7 November 2011 (UTC)
- [View source↑]
- [History↑]
Contents
| Thread title | Replies | Last modified |
|---|---|---|
| A 1% gain in APS just like that? | 6 | 19:24, 3 November 2011 |
(The only diff is adjusting the rolling depth of my gun stats.)
Wow. Did you increase or decrease it? If the only change was disabling your VG, I find d vs e even more remarkable...
Tbh, when I disabled the VG I was stupid and also added rolling to the main (and now only) gun. I also set the depth param for Paul's rollingAverage function to below 0.7. The next version had the param set to 10. And the one I just uploaded has it set to 50.
You do not have permission to edit this page, for the following reasons:
You can view and copy the source of this page.
Return to Thread:Talk:CassiusClay/A 1% gain in APS just like that?/reply (4).
Darkcanuck's NN guns are VG-less. I think he basically sums the output of two different networks (if Gaff/Targeting is still accurate).
Sorry you're having trouble with the LiquidThreads. I've hit problems here and there, but rarely. Maybe I'll check for MediaWiki/LiquidThreads version updates and hope for some bug fixes.
That's what I'm going for too. CrowdBee =)
