Win11 vs XP - simulation results

[Mitch-Dolphins]

I simulate games on both a Win11 and XP cpu. I've set both with Win98/ME compatibility mode.

Oddly, I've observed over the past couple of seasons that the simulation results for games using Win11 seem more variable vs the results using XP.

Has anyone else had a similar observation?

[Brian-Broncos]

Could this be the games not repeating after 50 games or so, like has been seen before?

I don't recall if that always happens or if it's based on how the simulation is setup - but maybe it's based on OS. I just remember Rich saying that it can somehow be prevented by changing anything in a profile or gameplan.

[Dean-Atlanta]

I sim using VMs that using Win 10. Using same PPPs and teams I see some variations from one VM to another but I think the explanation for this is that it's the random factors bult into FBPro98 at work. I think this is why in one sim the "FloA" play may outperm another PML play but not in the next sim, using sa,e PPPs and teams.

How are you duplicating or creating 100 or more game files to do your simming?

[Mitch-Dolphins]

Here's an example:

I create a game and generate 300 copies of it:

- Win11 - I have a 1.5 to 1 TD ratio vs my opponent
- XP - I have a 2.5 to 1 TD ratio vs my opponent

Same teams, same PPPs, same weather...

[Steve-San Diego]

So the real question is ... which OS does Rich use?

[Brian-Broncos]

I create a game and generate 300 copies of it

Is that enough? I'd think you'd need to be the 10s of 1000s of times to really know if there is a difference, and even that might not be enough to know for sure.

[Rich-League Officer]

So the real question is ... which OS does Rich use?

For a long time, I was playing all the games on Win7p, then last few seasons it was a mix of Win7p & XP.
But, sometime last season I had to abandon the older ones and all games are played on 2 Win11 machines.

I was aware of this issue with Mitch and it's hard to explain why this would be.
300 is not really enough unless each and every time it happens.
If I was just doing one test, I think I'd play 3000 and see how different it is.

Now you all know, games are on win11 so I would test on win11 if you can.

[Mitch-Dolphins]

What I shared was an example of what I'm seeing. I've likely done nearly 24,000 sims over the past couple of seasons in which I see roughly a 1 TD higher ratio with XP vs Win11.

Who knows... I don't recall this noticeable difference in the past. However, it's caught my attention the past couple of seasons.

[Dean-Atlanta]

Here's an example:

I create a game and generate 300 copies of it:

- Win11 - I have a 1.5 to 1 TD ratio vs my opponent
- XP - I have a 2.5 to 1 TD ratio vs my opponent

Same teams, same PPPs, same weather...

That is a far greater difference than I noticed while preparing for my week 2 game vs Wash. I was doing finals adjustments on my PPP and simming 500 games each time (probably overkill) and trying various changes, so to try other things I duplicated the setup on a another VM and it seems like no matter what I tried, opne VM my team did better. I did a contreol test and simmed the exact same PPP on both and on one VM I think my team wom about 10 percent more games and mabye scored 2 points more per game.

After some PPP changes, it flipped, and my team performed slightly better on the other VM. I finally the PPP from the VM and sent it to Rich and Charlie.

I think the game sets some kind of random factor when an exhibition gameis created and saved, as we do this from simming. I wonder if there is another random factor set in each install of FBPro98 on each PC or VM, and maybe this explains the variation we're discussing. I just know that it's annoying when this "statistical noise" makes it tougher to guage whether changes have improved a PPP, much like getting on the scale every day if you're trying to lose weight is a mistake because of day-to-day fluctuation of body weight. But in simming for PPP improvement, we have "get one the scale" every step, to see if we're improving a PPP.

Another issue I've seen in simming is duplicate scores, and in some instance, they are the same exact game repeated 100 percent idiontical.

[Dean-Atlanta]

I create a game and generate 300 copies of it

Is that enough? I'd think you'd need to be the 10s of 1000s of times to really know if there is a difference, and even that might not be enough to know for sure.

I've had this very discussion with one coach. When I first started simmming I did 100-200 games simmed per game plan that I tested. I've done 300, 500 and even tried 1000 at times.

I think somewhere between 300 to 500 unique game files is the reasonalbe number to get good sim data.