New article by Alex Rousso - Return on Investment

[MM]

There's a new article on the website:

Return on Investment by Alex Rousso

[evelyn]

An interesting article but I don't like your method. When I did science we used to reject the best and the worst results as duds (outliers). I think it's important to do this with poker results as a single big win can completely distort the real picture. If we do this with Toby Lewis and Chris Moorman on their current Pokerstars OPR results we get ROIs of around 22% - 23% for both of them

[Alex B]

I think that particularly for live poker the conclusion could mention the value of knowing if you can see opponents errors in estimating whether you have a profitable win-rate; a qualitive rather than quantitative measure.

Even with online cash, I think that a few thousand hands and an expert judgement is probably as good as 200,000 considering the usefulness of recent information and changing conditions etc

I.e. don't write off an investment because the ROI is unmeasurable, if there is another way to see it is clearly good (but still unknown)


Edit - Also if I was teaching ROI I think you have redundant steps, just do 15k/11k=1.36=+36%, no need to average per entry.
_________________
http://www.alexbowler.com

[evelyn]

[pickleman]

yes, the point of the article was to come up with some quantitative analysis - no matter how rudimentary. I think it's a given that people can make (decent) qualitative judgements about how good they are and how bad a field is. Though interestingly, I think that many players - even good judges - tend to overestimate their ROI. Hence, actually doing some spade work on the subject to come up with some figures - however approximate - was my intention.

What I'd really like to do one day is ask absolutely everyone in a tournament what they think their ROI (in that particular tournament) is. Obviously the geometric mean of all the answers should equal ~0.9 (or -10%) once you've taken out the juice. Can't help thinking it would be slightly higher than that

Moreover, it's not just the bad players who are overestimating their ROI. Once again if you asked all the pros/good players in a tournament what they thought their ROI was in that tournament, I'd bet you the "amount left over" would be less than -100% per fish, which of course is impossible. Just sayin'
_________________
Twitter: PicklemanPoker
Pickleman articles: http://www.bluffeurope.com/poker-news.../pickleman-poker/

[pvas]

I think that some of this analysis is a bit off. These online players are playing all HSMTTs and probably showing a positive ROI in all of them. However, playing a $200 rebuy with a 1.05 edge and a $100 rebuy with a 1.1 edge, really does skew the ROIs when they are playing the daily $10, $20, $50, $100s and everything in between.

Obviously a winning player should play all of the tournaments they can with a positive ROI, as long as it doesn't lower hourly, which is the main thing but skipping $100 rebuys+ and the sunday $500s and Super Tuesdays would create a much much much bigger average ROI imo.

As someone, who has mainly been playing HSMTTs on Sundays only and Tuesdays I would never say that the ROI I have achieved is my true ROI because Sundays are much softer than weekdays for example.

Basically I think you need to filter for buy in sizes and although lots of people's SNs are blocked OPR is pretty good for tournament rankings and sharkscope too of course, with subscriptions you can filter these sorts of things. It would be interesting to see the players you analysed at different buy in ranges, although I realise the sample size would be even smaller.

Soft tournaments do exist and big ROIs are possible but like you I would hate to estimate them. 200% ROIs live doesn't seem a massive leap imo but it does online for someone who plays everything.

Didn't mean to post something that long but oh well

[pickleman]

This is an old article now, but I did something like that here with SNGs:

http://www.thehendonmob.com/alex_rous...e_lucky_than_good

It's easier with SNGs because the prize distribution isn't so massively skewed in favour of the winner, and people can play more of them, so the results settle quicker. I seem to remember that the ROI of good $100 (turbo) SNG players was around 7% but for e.g. $20 SNGs it was around 20-25%.

Yes, the same phenomenon will exist for MTTs but the samples really aren't big enough to start factoring for buy in. What I'd really be interested in is to factor for field size and look at the variation in sample ROIs for players with the same overall ROI. I'm pretty convinced that large field MTTs are too much of a gamble, as I analyse here:

http://www.thehendonmob.com/alex_rous..._much_of_a_gamble

In that sense I concur with Evelyn's point that outliers skew the results (an extra win in a 2k runner event over 10k tournaments will have a marked effect on the ROI) but you can't just remove the outliers because they're an absolutely germane part of the dataset. Removing those outliers is to essentially remove the money from the prize pool (and therefore reducing the whole dataset's ROI. Still, there are ways to put that money back, but that's for another article, perhaps.
_________________
Twitter: PicklemanPoker
Pickleman articles: http://www.bluffeurope.com/poker-news.../pickleman-poker/

[MetalBox]

I feel your approach is pretty un-scientific and just really your own guess. You seem to have a looked at a few of the top scoring players and found their roi for some period and by some way not explained decided that if we reduce this by some fraction this is a reasonable estimate for a 'good' high stakes player. The approach you have picked is very easy to argue against, there is great bias in the data and as far as I can tell it is really only your intuition.

I am not a statistician or mathematician but I am sure there will be ways of doing a much better job with a more scientific approach and there is plenty of data available to do it with. Statisticians do a great job often with very little data, they have to in fields like medicine where you want to only 'hurt' the least amount of people when doing drug trials and yet they still need to know how accurate the results are. Alas, I personally do not know enough to do a job of producing a method of estimating roi that is accurate and believable but non the less I will try.

To overcome the bias (to some extent) when using the top figures from a database of results you could select the top 100 players from last years results. This should produce a group of players that we could say were 'good', some would be mediocre and lucky and some great but unlucky but it would be easy to argue that the average across all of them is a 'good' player.
Next from all the tracked games these 'good' players have played in the current year, (a different independent data set from the one we used to produce the 'good' player set).
For each individual game played above a certain cost calculate the roi. (this will smooth out any bias in results from one person that plays many $100 tourneys and few $1000 ones, so we should get a reasonable overall view from both easy $100 and harder field $1000 ones)

Use this list of roi's to find the mean roi and also the 'confidence interval' CI, 90 or 95% CI is often used, on these. This is quite easy with many math packages.

This should give a reasonable +/- roi range for our group of 'good' players. We should be able to say that 19 times out of 20 a good player's roi will be mean 'x' + or minus 'y' for the near future.

I don't have much of a life but I am afraid I am not quite lonely enough to do this work. Alex, are you busy this week? Just joking, it would be good if somebody with more knowledge than me did produce an achievable roi score chart for MTT players, it is such a hard occupation and such a figure is useful to know before trying to become an MTT professional.

edit: this is just for the money earned from the games and ignores the likes of rakeback and sponsorship on the value of playing HSMTT's.
Perhaps somebody could persuade one of the online sites to produce an average roi value for say the top 2% of players from the previous year tracked into the current year - I suspect they would only publish this if the result was pretty high. They will also have statisticians and so should be able to find a better method. Also, just producing a figure without eplaining how it was obtained would be nearly as useless as Alex's method

[evelyn]

[MetalBox]

[Pizzicato]

A few comments...

Firstly, there are likely to always be differences between Stars and FTP ROI's because the structures of the tournaments were actually pretty different and had their own peculiarities. Also the differences in field sizes will have had a decent effect as, in general, the smaller the field sizes the smaller the attainable ROI's are going to be.

Secondly...

Pokerstars HS tournaments (with a few exceptions like Sunday Majors and SCOOP/WCOOP) do not have thousands of runners... Most of the HS tournaments actually have more like 150 - 400 runners with some having less than that even. I can only think of 1 PS high stakes daily tournament that regularly gets more than 400 runners and thats The Big 162 which generally has a higher fish - reg ratio anyway.

Thirdly...

Those HS tournaments in questions are generally amongst the toughest online fields going. Its not like playing a 180 man with with 100 regs and 80 fish... Its more akin to playing a 180 man with 150 regs and 30 fish...

Fourthly...

Once you get away from the high stakes tournaments and you come down to more mid - low stakes tournaments, possible ROI's jump because of the fact there is a higher fish - reg ratio and the field sizes start increasing. The more fish you have the chance to outplay, the higher the possible ROI's are. 100% ROI's are still possible in low - midstakes tournaments though the variance often increases due to massively increased field sizes if you were to compare like to like field size wise this would shine through like a lighthouse on a clear night.

Fifthly (hmm maybe more than 'a few' comments after all lol)...

Sample sizes... You are most definately correct though I am unsure if you even realise how correct lol. 10,000 games is where most people start to consider 180 man turbo ROI to start getting somewhat close to a reliable ballpark figure. However I found that even after that amount of games a hot or cold run could see my roi swing by a % point or even 3 and the real line is probably closer to 20,000 games. That being said its still going to be on the low side for the HS tournaments with 150 - 400 runners (especially considering the higher variance of there being more +ev players in the field) and completely absurd once you get into mid - low stakes tournaments where average field sizes can jump to 2000 - 3000+.

I would be very interested though if it would be possible to work out at what point the required sample size vs field size for true expected roi would begin to converge...

Anyway... enough rambling for now... I need to get some sleep so I can ride the variance wave tonight...

[pickleman]

[evelyn]

[MetalBox]

[evelyn]