[OUTDATED] The Return of FGei (for v1.32x)

[rob77dp]

Pella hasn't answered yet. :/ Maybe she's offline?

Pella is male and the little white box by the username indicates offline (green box indicates online).  It has only been a couple days and I'm not sure you'll get a PM response, but rather the stats will just be included in the study/stats.

Lastly, I don't think this is a valid situation for a double post.  Instead, use the "Modify" function in the top right area of your prior post and put your additional text as an addendum to your previous post.  (Forum users nod "thanks", Submachine.)

[Zso_Zso]

Question about a small detail of the calculation:

Does the formula in the spreadsheet use an average spin-chance for all FG opponents, or does it use the specific chance per FG ?

They do differ in a range of ~30%-56% and since some decks work better against some FGs while different decks will favor different FGs, it would make sense to use the specific numbers rather than an overall average for more accurate evaluation of deck performance.

There is an old thread about the specific win-rates:
http://elementscommunity.org/forum/index.php/topic,20376.msg420980.html

I guess, I could check in the spreadsheet, but its easier to ask -- sorry for my laziness 

[ColorlessGreen]

Question about a small detail of the calculation:

Does the formula in the spreadsheet use an average spin-chance for all FG opponents, or does it use the specific chance per FG ?

They do differ in a range of ~30%-56% and since some decks work better against some FGs while different decks will favor different FGs, it would make sense to use the specific numbers rather than an overall average for more accurate evaluation of deck performance.

There is an old thread about the specific win-rates:
http://elementscommunity.org/forum/index.php/topic,20376.msg420980.html

I guess, I could check in the spreadsheet, but its easier to ask -- sorry for my laziness 

Per discussion in the UEI thread, the current GDoc FGEI spreadsheet uses an average. I suggested in UEI that the FGEI GDoc spreadsheet be converted to use actual spin rates. I don't believe this has been done yet (or at least, if it has, I haven't seen it mentioned that it's been done yet).

TBH I would consider any result of the GDoc spreadsheet to be a work in progress right now. There are other assumptions that have changed for the GDoc spreadsheet from what has traditionally been used in FGEI.

edit: To clarify, I think some of the changed assumptions are good changes, but they are different than what FGEI has meant in the past.

[Keeps]

I know the number of cards being won is added in, it is possible to use those values to calculate spin chance and it is sort of already included if you have enough tests; however, if the numbers are already calculated than I would suggest that is used if it's not already.  (I don't know if it is or isn't)

[ColorlessGreen]

Spin chance (both average and individual) is known. The GDoc spreadsheet (as of last time I heard anything about it) was using average spin chance rather than individual spin chance. It's not a big deal overall but using average instead of individual means that a hypothetical deck that beats divine glory 90% of the time and rainbow 10% of the time is more profitable than another hypothetical deck which beats rainbow 90% of the time and divine glory 10% of the time but is otherwise identical.

In real world situations, it makes a tiny difference in total money but doesn't really tend to change the rankings much if at all. It would still be better to use the individual chances rather than the average chance, though, since IMO pretty much anything that improves accuracy is worth doing. Previous FGEI-calculating spreadsheets have used individual chances.

[Keeps]

Yea, I think this time around it might though, last i looked at the spread sheet.  We have 3 top contenders...  although two are showing 'new' and one 'unconfirmed'
They are all mid to high 8ks.  A 10% or 20% change could make a difference especially in voodoo's cases versus the other two.  Since voodoo has a more mixed win loss ratio versus malignant and pdials have specific FGs. 

[ColorlessGreen]

A 10% or 20% change could make a difference especially in voodoo's cases versus the other two. 

A 10-20% change would be huge. I'm pretty much just speculating here but I'd honestly be surprised if we even hit a full 1% change on average, though I guess I could see certain individual decks getting more like 2%.

To repeat, though, still very much worth doing. (Particularly since now I'm curious about how much it'd actually change the results.)

edit: Welp, turns out I was wrong. I just did a quick test using the data for PDials-32. The average amount of money won (from spins only) for one game against every FG using average spin chances was 7834.76, and the average amount of money for one game against every FG using actual spin chances was 7008.67. Note that this is the most exaggerated this is going to get (as it's not factoring in anything but spin chances, and 29 games is a good bit more than an hour with pdials in almost all cases), but it's still IMO proof that it's significant (and probably in the 5-7% range I'd guess for actual change to FGEI). So, yeah, please fix this.

Test details:

Spoiler for Hidden:

Win/Loss/Skip totals pulled from GDoc; win rate per FG calculated.
Average profit per game per FG calculated as 1177 * WR * Spin Chance. 1177 was noted by Pella as the value used for an upped card during discussion in the UEI thread.
Spin chance = 45.66% for average. Noted by Pella as value used during discussion in UEI thread.
Spin chance = taken from here for actuals.

[Keeps]

While we are on the topic
Do you know if it's subtracting out the cost of the loss?

[ColorlessGreen]

While we are on the topic
Do you know if it's subtracting out the cost of the loss?

I haven't audited the GDoc but I'm almost certain it is because the results from the GDoc aren't very far off of results from previous known-functional tests.

Also, I still think someone (who isn't me) should just pull the raw data from the GDoc and plug it into statmasta and compare.

[Leodip]

Pella himself said that it uses the same exact formulas from the Statmasta, hopefully it should give the same results.

[ColorlessGreen]

Pella himself said that it uses the same exact formulas from the Statmasta, hopefully it should give the same results.

Where did he say that? Cause he said in the UEI thread (see this post, including quoted materials) that he was using average spin chances instead of individual spin chances, that he was using a different average spin chance which he self-calculated, and that he was using a different upped card sell value which he self-calculated, and a self-decided skip time (which IIRC is different but I haven't double-checked).

If the results of both the GDoc and StatMasta matched exactly given the same input data, I'd be quite happy.

[Leodip]

Sorry, I got it wrong, he was talking about another matter in the post I remembered.