Confidence Calculator

[AleoMagus]

As I mentioned in another thread, This week I've been meaning to put together a handly little confidence calculator and today I finished it.

It is on the same site as my 'Universal' spreadsheet

http://www.aleomagus.freeservers.com/Spreadsheet

It is the Confidence Calculator file and is an excel spreadsheet

It's pretty interesting to fiddle around with different numbers on this thing. Gives a good idea of just what kind of swings one can actually expect and what kind of samples are really all that meaningful.

As ever, I don't even claim to be an expert so I'd love feedback from those who know more about stats than I do.

Actually I'd appreciate any feedback or suggestions

Regards All,
Brad S

Oh yeah - some have difficulty downloading these things. My only suggestion if you do have trouble is to not link to it, but instead type in the site manually and go to it manually. Something about using 2+2 bandwidth to download from freeservers. Maybe someone else can explain this better.

[AleoMagus]

hmmmm...

I just integrated this calculator into my 'universal' spreadsheet and made a couple other changes

I made it so the Profit graph title generates correctly instead of saying 20+2 automatically (thanks Durron)

I made it normally include 300 sngs instead of 100

Just one problem - Freeservers is telling me that the new file size is too big to upload. Does anyone know where I can put a file like this, or how I go about doing it?. While I am at it, how do I make a nice link on my main page instead of directing everyone to those ugly folders.

Please do not PM me asking for me to email this new version. I'd rather just get it up on a site.

On other thing...

I am not actually very possessive about anything I contribute on this forum, so if anyone wants to muck around with any of it and repost it (better hopefully), I have no objections at all. I do appreciate being sent a PM or email though and I might be upset if I saw I spiced up version on ebay (unless I got a cut) /images/graemlins/grin.gif. Seriously though, I doubt it's particularly valuable to anyone but people who come here and get it for free anyways so feel free to expand.

Lord knows that is all I am doing anyways thanks to the stats lessons of posters like Bozeman and the other math Gurus.

Regards
Brad S

[durron597]

This calculator is quite uber. Well done. I find it interesting (after putting fun numbers in there) that you need to have about 1000 SnGs before you can be 90% confident to within $1 (for $10+1s). I guess it will be awhile for me...

Anyway, I don't know if Mason accepts PMs, but maybe you could ask him to put it up on 2+2?

[Tosh]

It looks impressive, though I haven't got a f***** clue how it works. /images/graemlins/crazy.gif /images/graemlins/laugh.gif

[AleoMagus]

The thing which I find most sobering is the normal standard deviation associated with small future samples (like your next 50-100 sngs).

Most good players with typical standard deviations can only be certain they will get anywhere from break even results to double their usual expectation.

This is to say that if you are accustomed to long term ROI in the neighborhood of 30%-40%, in your next 50-100 sngs, you could get anywhere from small negative ROI's all the way to 80% or more. It is that swingy. Kinda makes the '100 SNG challenge' seem a little silly.

Even in long term projections of 1000 or more, the difference in potential expectations comes in thousands of dollars.

Imagine saying this - "I am a solid 40% ROI player, and in my next 1000 sngs I suppose I'll make somewhere between 3000 to 5000 dollars"

A more imprecise figure than many probably hope for, I'll bet.

Regards
Brad S

[zephyr]

Thanks,

I was just coming on to post a question about some large swings I've been having of late. Needless to say, your post, and calculator answered my question wonderfully, and very precisely.

Thanks for the confidence boost,

Zephyr

[Barrett's Last Privateer]

Brad,

Let me start by saying I think your contributions to the forum are excellent, and so the following is only meant to (hopefully) improve upon a great foundation laid by yourself, but I think your calculation of standard deviation of mean profit is slightly incorrect.

I had a play around with some numbers, and it seems your formula takes the squared differences of payout from the buy-in ($11), rather than the squared differences of profit from the mean profit per SNG ($4). Also, you haven't included the non-cash results in your calculation.

When I recalculated the figure using the above chnages i.e.

SQRT( (1/100) * ( 15*(39-4)^2 + 15*(19-4)^2 + 15*(9-4)^2 + 55*(-11-4)^2 ) ), I got a standard deviation of $18.57.

Another possible chnage would be to use 99 rather than 100 as the denominator in the equation, which chnages the results above to a s.d. of $18.67.

Now obivouly these numbers are in fact very close to the figure in your spreadsheet ($17.16), so the difference may not be significant, and sorry if this post seems picky. It may be (I haven't checked) that the differences betweeen your calc and mine are bigger for different distributions of results, and it may also be, of course, that I am wrong, everyone should feel free to flame away if I am. /images/graemlins/smile.gif

Cheers

[AleoMagus]

[ QUOTE ]
Also, you haven't included the non-cash results in your calculation.


[/ QUOTE ]

Oops! Thank you. I have fixed this. I thought the SD looked a little low (I usually suggest that 2nd place net profits are a good estimate for SD for those who simply do not want to figure it out)

As far as your equation for SD, you are likely correct, but this is not the equation I first learned for calculating SNG Standard deviation.

Take a look at my fixed equation and let me know if you still believe it is incorrect.

In truth, it is a small matter to fix it, but I'd like more feedback on my current methodology first. As I said, I am not a stats expert and I'm not sure I can expertly say one is better than the other.

Thanks
Brad S

Note to all - I have taken the 'universal' spreadsheet off my site, but this should only be temporary as I will soon figure out a way to get the newer version online.