Final Table only would be awesome too. You make lots of final tables when you play 18 ,27 or 45 man sngs.


Equal-stacks assumption/speed-up is brilliant too offcourse.

Well in MT SnG PT3 knows about the number of players but in general MTT's it doesn't unless a complete summary has been loaded in (as in Full Tilt summary log). Also, calculating ICM for a large tourney is very expensive (takes a long time) so I don't think I'll be doing anything beyond the final table.

Does this work with the trial version of PT3? It only gives me a blank graph. I am not sure what I am doing wrong.

This program is the sickness for curbing tilt. I can just sit back and say "I played correctly" after looking at this chart. Would totally buy it

Agreed, great to see you're "running bad", however we won't actually know to what extent till there is actual $ graphed in. Then it will be epiccccccc, I'm checking this thread like 5times an hour in the hope it's out, when it does get released it's going to be like christmas imo.

100% agree

Every hour I'm like I was as a kid on xmas day getting up at 6AM to open his stocking.
Seriously, software is sick good already though, my BB swongs are very sick though over 6k all ins graphed.

New version contains a couple of bug fixes and a new ICM line in the graph:

• Was not taking into account dead cards in side pots. I found this bug when I was all in with a set of 3's against 2 players holding QQ each This requires a full recalculation to effect old hands.
• App now looks at 64 bit registry for PT3 install dir.

Some notes regarding ICM.

• I can only calculate ICM for STT's and I need to know their prize structure. For this reason, I only calculate ICM for "SnG (1 Table)" and "SnG (6 Max)" (Now that I think about it, I could also add (2 Max) I guess). I assume prize structure to be 50/30/20 for full table and 65/35 for 6 max. I'm not sure if this includes Turbo STT's or not.
• The ICM graph displays the difference only between the expected ICM and the actual ICM for the given pot. For example, in a $5 6 max STT, the prize pool is $30 and 1'st place takes home 65% or $21.50, if your expected ICM is $8 and your actual ICM is $0 because you lost and busted out, your ICM value on the graph will be -$8.
• I haven't added columns for ICM in the hands list yet.
• I haven't tested much so I need you guys to keep me honest. Don't take results for granted. Help me find bugs.
• The first time you run this, it'll detect that you haven't done ICM yet and recalculate all your hands. This will take a while so be patient.
• I already see some interesting results for example, i could have a hand where I am all in as a 30% underdog but my ICM luck chart drops a lot. I was surprised by this at first but then I realized that it was a bubble hand and even my 30% equity in the pot gave me a high expected ICM value by doubling me into 2'nd place while loosing the hand dropped me to zero because I was knocked out.

Download Link

If you test it against my ICM luck code and don't get the same results then you know for sure one of us has messed up and can then work through the maths manually for the hand to see what the problem is.

PS: Don't expect the results to be exactly the same though:

A) For preflop 2-way and 3-way all-ins I use a 169x169 and 169x169x169 lookup table for the outcome chances so flush dominance isn't accounted for (although experimentation shows this makes almost no difference).
B) For 4-way (or more) preflop all-ins and post-flop all-in I use Monte Carlo simulation to generate the outcome chances. Since I use 20k roll-outs though it will be very close to the result if I had of enumerated (just never had time to implement enumeration yet).

Even so they should be very close and any major errors should show up as significant differences in output.

Juk

Thanks. I don't know where your code is though...

Here's what I do:
I found some code online that calculates ICM recursively given stack sizes and prize structure.
I already calculated expected and actual chip counts for all in players so I added stack sizes for players who folded (expect & actual are equal)
I calculate ICM for hero with expected stack sizes and actual stack sizes and compare them.
I treat each pot separately so a 3 way hand will have 2 values, 1 for the 3 way main pot and 1 for the heads up side pot.

Has this worked for you in previous versions or is this the first attempt? Do you have any filters defined?

First attempt but I got it to work. I think it was on the PT end. Thanks.

The lastest version is here: http://www.jukofyork.com/Juks_BackTe...uck_v1_03b.rar

If you look in the "SourceCode" subfolder in that RAR file then you will see the way I did it.

I'm not quite sure how you are doing it, but one thing to be clear on is that you can't just take the hero's "average stack" at the end of the hand and pass it through the ICM function to convert chipEV into $EV. Using the "average stack" is fine for working with chipEV, but because the ICM function is non-linear you have to work with each of the outcomes separately. For example, to calculate the $EV_luck for a 2-way all-in you would need to work out:

$EV_luck = P(win)*$EQ_win + P(tie)*$EQ_tie + P(lose)*$EQ_lose - $EQ_actual

and for a 3-way all-in you will need to work out:

$EV_luck = P(outcome 1)*$EQ_outcome_1 + P(outcome 2)*$EQ_outcome_2 + .. + P(outcome 13)*$EQ_outcome_13 - $EQ_actual

(you use the ICM function to estimate the $ equities, $EQ_x).

It gets much more messy with more than 3 players (I forget how many outcomes there are for 4; 80ish IIRC) so that's why I (re-)wrote my code the way I did (see the "outcome generator" class).

Juk

w00t.
Running now - I've got about 6k all in hands from before so reasonable sample size I reckon.

One question - I've got $6's mixed in with some $11s, will it realise this and do the proper calculations?

Cheers and w00t this is exciting

Ok it finished re evaluating everything and now I just get a blank graph. No writing or anything, just a X and Y axis and "All Dates (T35,407) on the top.
"View>Big blinds" Also gives same thing, no lines plotted, but I see "All Dates (683BB)".
Strange.

EDIT: Right clicked graph and "set scale to default" fixed it.
So now there is an added 4th line on the graph - ICM. However at hand 0 it starts at 3,000BB, and is horizontal over the 6k hands I have recorded. Like perfectly straight.

Edit 2: This possibly could be because all of my hhs are Turbo stts? I see you mentioned something similar earlier?

which sites does this software work with? does it work for non FTP/PS? any chance you could add a 70/30 6 max structure payout for some euro sites op?

are there any plans to get this to work with holdemmanager?

Yeah, I guess you have a pount there. Not sure I really wanna go down this road. Too much work
I'll take a look at ypur code tho.

The ICM graph uses the right axis so it's probably stuck at zero, not 3000. Yeah, I can't do turbo's right now.

Ah ok, do you need any hhs emailing to you to help?
Cheers

I've got error. Here it is.

Guys, help to resolve it please.

Can someone help interpret my ICM line? It appears up-and-down jagged until about 650 hands, then falls like a stone. Does this indicate a horrible run of ICM luck? (the "Difference" line is much more stable, by comparison).

I think it indicates that there is something wrong with what I'm calculating.

If my database would be of any help in your debugging, PM me and I'll figure out a way to back it up and send it to you (I'm a software developer, but not a Postgres guy, so I'll need to figure this out, but would be happy to do so if it helps you).

Would it be possible to add a filter which graphs only hands which took place heads-up? I feel like I'm running signficantly below expectation heads-up.

Also, have you given any additional consideration to the idea of generating a graph in terms of strictly EV (no regard for tourney chips, bb's, icm etc)?

Thanks so much for your hard work!