Phil,

I'm taking suggestions from an experienced developer like you very serious.

I was kinda afraid connecting directly to PT because I've watched the development of PokerEV and you must have had a good reason building your own parser.

However, I don't give up yet since I only have to sort out pre-flop action whereas PokerEV has to deal with all streets. Just trying to make it as convinient to people as possible and import from existing databases is a nice feature. Sooner or later I will have to build my own parser anyway.

As for the forced internet connection: I've never seen it that way, but your objections make a lot of sense. I just had a hard time getting the right time-zone and daylight savings / standard time from Windows, but non US (EST) users can still set it manually of course. So, required internet access will be gone with the next release.

Thanks for you post!

Just to be clear on the net-connect issue, I did a cursory check with netstat during execution and it does appear to only connect to timeanddate.com. But I fully support Phil's comments there, you have to demonstrate that users can trust you, and that means either open-source or cut out all the auto net-connects. Do that and you could have a big success on your hands! Keep it up, a lot of potential here. Been looking for something exactly like this so am very excited about the prospects.

Yeah, a lot of poker players are paranoid about giving programs internet access, especially after rbcalc.exe. People are even complaining about PT3 requiring internet access, and they're as trustworthy as you'll get.

As for accuracy, PT3 is out very soon and will likely fix most of these issues...should make your life a lot easier. PT2 accuracy is really good for FT and Stars (at least for ring games) which is most of your customer base anyway. Giving an option to limit imports to those sites will get rid of a lot of accuracy complaints and weird looking graphs.

Like I said good luck, innovation in poker software is really needed.

Great software! I hope it proves to be accurate because it has made me feel much better about my recent break even run and downswing. I'm in the process of importing more tournaments into an access database to see how it goes over a larger sample.


Have you considered including in analysis all in pre flop situations not involving hero? For example when a shortstack doubles up in an all-in with somebody else reducing our equity. I imagine the effect would be quite small compared to the equity difference when hero is involved but it would be interesting to see how small if it was easily implimented.


I am having problems with using the Date filter. When I click "Pick Date" a calendar to pick from flickers up but rapidly disappears before I have time to select anything. It seems to usually pick todays date, but if I double click very quickly I can select a date as long as it is positioned on the calendar above the "Pick Date" menu item.

I tried unplugging both mouse and keyboard (seperately) to check I didn't have a problem with a button stuck down but the calendar still disappears.

I am using vista if that makes any difference. Does anyone else have this problem?



PS My vote is for postgresql support before ROI calculation

Nice work on the update, I imported 1800 sngs to test and was -$62, very close to expectation. Here's some random suggestions to keep you busy:

- Importing from PT database isn't actually convenient, most players want to use this program to analyze a downswing/heater so they only want to import a sample of hhs not their whole database. (stars summaries are also a pain)

- At the moment "Luck" is simply a measure of equity difference, so it wont converge for large samples. Just take (["Luck"]*[#players]*[buy-in without fee])/([#sngs]*[buy-in + fee]), use this to describe the luck in a sample, this is the same as the effect on your ROI.

- It's possible to analyze postflop all-ins in the same way you are analyzing preflop all-ins. More hands being analyzed = better accuracy.

That's another good idea, when the shorter stack keeps sucking out the bubble it can have a big impact on your equity. The problem is it may increase import times significantly while only improving accuracy slightly.

IFold,

I was interested in your comment here. Are you saying that all things equal, the luck number should not go to zero in the long run? Perhaps I'm just misunderstanding what you mean by "converge" here.

Oh and I agree that a non-PT option to look at hands would be nice.

wiki - "For example, we may see 520 heads after 1000 flips and 5096 heads after 10000 flips. While the average has moved from .52 to .5096, closer to the true 50%, the total difference has increased from 20 to 96."

At the moment "Luck" is the total difference not the average. The average converges but the difference from expectation doesn't.

Ok thx for the explanation. Makes perfect sense.

New BETA release:

SnG Luck Analyzer v0.1.6

Party should be fixed.

• New handling of date and time. Internet access is not required anymore. Instead, the date filter can be set to "Day of current Hand" and "> current Hand" if a hand is selected. All hands are now saved with the actual poker-site time.
• Fixed a bug with Party & Afiliates. The amount of luck was generally too high since returned pre-flop bets haven't been captured.
• Party & Afiliates: SnG LA tries to split buy-in and fee if Pokertracker has wrong tournament summaries.

Also, I hope Vista users can now see the date picker.

What abouyt TurneyManager support bodypull, a lot of sng-players use that.

And that way I dont have to ask Stas to email me tournamentsummaries.

Just remembered you need a post-flop equity calculator to do this since you can't just look up a pre-calculated table, that makes it a lot harder to do.

I am still experiencing the same problem with Vista using the new patch.

The "> current hand" option is a good alternative for now though.

Actually I'm thinking about something else. Why not auto-complete the Pockertracker database since all needed informations are available?

Let's say the problem is to distinguish between STT and MTT.

• If you are sitting not at table 1

-> it has to be a MTT

• If you are sitting at table 1 AND
• If there is no other table with the same tourney number AND
• If you are getting down to 3 players without having new players joined the table since the very first hand

-> it has to be a STT

What if you are coincidently sitting at table 1 in a MTT AND bust out before new players join AND before it comes down to 3 players?

Well, we don't care. If you haven't played under the same tournament number at a different table it has to be either a STT or a MTT where you didn't finish in the money. So even if we accidentely call this tournament STT instead of MTT it doesn't change your ROI at all.

Generaly only tournaments without summary would be touched anyway.

The only thing which bothers me: If it is that simple, why did nobobdy do it yet and why doesn't Pokertracker do it with a build-in function?

Please correct me if I'm wrong.

I dont know.

Lets try it.

I have just finished importing 5028 $16s and 261 $27s into an access database, and SnG LA says that I am -$2781.77 in luck at $16s and -$1463.65 in $27s. These are all played on pokerstars.


Just looking at the larger sample of $16 tournaments this seems to be quite a large deviation over so many tournaments. It is equivalent to a ~3.5% ROI difference. ROI Simulator says that there is an 8.2% chance of being down >3% and a 3% chance of being down >4%, so perhaps about a 5% chance of being -3.5%. I would think that the variance in all in results would be much smaller than the overall variance due to so many other factors of luck, so the chances of being this down on all in luck would be <5%. I guess it's not beyond possibility that I've just been that unlucky depending on how much of the variance in STTs is accounted for by SnG LA. I have little knowledge of statistics so appologies if any terms used above are incorrect or if it doesn't make actual sense. I hope you see what I mean though. Thoughts anyone?


A further observation is that when filtering by Push/Call I get -$27.67 over 3254 hands filtered by Action:Call and -$2754.09 over 11993 hands when filtered by Action:Push. It seems quite possible that there is an error in the calculation for Pushes but not Calls. I will have a play with the other filters to see if I can narrow down where the problem is if there is one, but when adding any more filters the sample size may be too small to see any patterns.


Has anyone else imported a largeish sample of stars tournaments? Can you check the results when filtered by Pushes and Calls?

I have just 350 tournaments, because I had to ask Stars for every single summary.

I went insane yes manually adding the emails.

But I have a + by Push which is about as big as the - by Call.

just updated to the new version from the initial. Is there a way to still scan a fold of HH's? For me this is preferred to pokertracker. I looked through the quickstart but didn't see anything. If not, can this be added back as an option?

Hey,

I got two problems.

First, I need almost 50 minutes to import only 2000 SNGs. Is that normal?

Second, I think the new version still doesn'T work with PP. See Screeny:



but still good work!!

matic

I have now imported about 1050 $6.50s as well and got much more normal results of +$33.3 Push and -$90.01 Call. Maybe I was just that unlucky at $16s and $27s so far.

agree

Not at the moment but maybe in the future.

This looks like an import from a previous version since the buy-in isn't split in real buy-in and fee. Have you reimported your hands with v0.1.6?

Dave, I'm watching your posts with great interest. Unfortunetely there isn't much I can do to verify besides double-checking some hands manually. Would be astonishing to have such a long streak over more than 5k tournaments. But who knows? As of today we don't have any reference about what is normal.

Thanks for your updates and keep them coming!

I think you are correct when you say that the variance should be lower than the total variance for your sample ROI. This is because there are many independent factors that contribute to overall variance, since they are independent the sum of each aspect of the variance should equal the total variance. (although you take the square root of the sum of the squares when adding standard deviations, which is probably what we are talking about here)

About the pushing/calling stats - you have a lot more pushing hands (12k) compared to calling hands (3k), you already knew you were at -$2800 in total, (assuming the equity in the pot is a similar for pushing/calling) that means you would expect roughly -$2800*3/15 = -$560 for calling and -$2800*12/15 = -$2240 for pushing. So your stats aren't really as far from expectation as you might think.

Here's my stats for 1800 $16s at stars (thats all I have summaries for):


Maybe you just run horrible...

I've downloaded the latest release and imported 5815 Party SnGs. Now, I know that I run gut-wrenchingly horrible 24/7 but can this really be accurate?

Luck -2301.1%
Bucks -8725.85$



I don't know how to use ROI simulator and I'm not a maths guy, so what are the chances of this?

Interesting post.

I'm not an statistic expert either and until today we don't have any references about what a common long-term deviation could be.

On the other hand, if you see it that way: in average your bad-luck is "only" -0.39% per tournament or -0.15% per hand.

The big question is: how many tournaments do you have to play to get back to zero? How many coins do you have to flip to get 50.00% head and 50.00% tail?

I'm using every opportunity to verify, recalculate and double-check hands but still, I cannot guarantee that the math is 100% correct.

However, statistic and propabilty is a weird thing...

http://www.youtube.com/watch?v=xMSjn...rsist_locale=1