Yield or Units Profit?

[ArtMaster]

Hello,

Theoretical Question here...............

If after 10 years there was a system which showed to be profitable, not every year but overall;

If there were a few variations of this system - one which made 20% Yield and a profit of say 60 Units, then another which made 10% Yield but 150 Units then which should be persued?

Also to help out, System A in all that time has made 400 picks say, whilst system 2 has made nearer 2000.

Should you go with the one with most profit, and double up when the first system made the same pick?

Obviously the second has many many more picks, much more incorrect but many more correct also. Since after 10 years you could have a decent amount of confidence in both variations of the system, would you be best off going with the one which showed significantly more profit or the one which showed a significantly higher yield?

Hope this makes sense!

[aliando]

always look at the yield - units returned is simply based upon units staked. if one system offers you 10% return of your investment and another offers you 20%, then the second system will always give you more money.

others may explain this better!

[Dices]

In brief, since English isn't my best:
When you compare two systems you sould look besides profit-yield and at other one major factor, the risk factor (depens on amount of bets,odds,etc).
I personally would go for the points gained (is this what all we are looking for ? ) assuming that the risk is at 'acceptable' level.
Dices

[phantomDude]

Originally Posted by aliando

always look at the yield - units returned is simply based upon units staked. if one system offers you 10% return of your investment and another offers you 20%, then the second system will always give you more money.

others may explain this better!

You say that if one system gives a 10% yield, and a second system gives 20% yield, the second system will aways give more money. In the cases that ArtMaster gives, over a fixed period of time the system with the lower yield gives more money though!

If we're comparing the performance of two systems where one sytem has been tested over a dataset five times larger than that which the other system was tested on, we're not comparing like with like. The first system gives a 20% yield over 400 bets, but you can't extrapolate from that and assume that it will perform similarly over 2000 bets. I replied to someone elses question in another thread where I illustrated that the performance of a betting strategy can fluctuate wildly over batches of 300 bets:

Serie Beeeeeee, i love you

Personally, I would go for the system that has been tested over a larger number of bets, because it's *far* more likely to be in profit because it's a system that capitalises on some inefficiency in the markets, as opposed to a system that just "got lucky".

The original posters question interested me, it's one I've often asked myself. In order to answer it, I wrote a program that queried a database of football matches, selected a number of fixtures at random from the database (the exact number of fixtures was specified by me), and analysed how much I would win if I bet on the home team in each of those fixtures. *However*, as I realise that this in itself is not a conclusive indicator of how likely you are to be in profit after 300/500/1000/10000 bets picked at random, I wrote the program in such a way as that it would perform this operation however many times I wanted to.

The algorithm for the program was:

1. Pick X games at random from the database (where X is a number specified by me)
2. Analyse how much I would win if I bet on the home team to win in every game.
3. Perform steps 1 and 2 Y times (where Y is a number specified by me)
4. Analyse how many times out of the Y iterations steps 1 and 2 resulted in a profit.

The point of this exercise was to determine how likely I was to be in profit after X numbers of bets merely by chance (remember, my bets are being picked at random), and not as a result of my system fluking it.

I ran this program last night using a database of 73399 matches. I ran it with the following values:

X=400,Y=1000: I picked 400 matches at random from the 73399 matches, betting on the hometeam to win, and taking note of my profit/loss. I did this 1000 times, each time keeping note of my profit/losses.

X=2000,Y=1000: I picked 2000 matches at random from the 73399 matches, betting on the hometeam to win, and taking note of my profit/loss. I did this 1000 times, each time keeping note of my profit/losses.

I am trying to estimate the probability that, after 400 bets and after 2000 bets, that I am in profit purely by chance. Remember that in the above process, I pick matches at random: there are no other selection criteria applied.

(I have images illustrating the below, I don't know how to insert images, or spreadsheets from my PC onto Punters Lounge, I'll put these up when (if!) I figure it out).

Out of the bets were we tracked the profit/loss after betting on the home team on 400 random teams 1000 times, I was in profit 18 times. My profit/losses ranged from +18.83 units to -97.55 units.

Out of the bets were we tracked the profit/loss after betting on the home team on 2000 random teams 1000 times, I was *never* in profit. My losses ranged from -74.65 units to -347.29 units.

This illustrates what everyone here knows: you may find yourself in profit over a small number of bets, but the more bets you place, the tougher it is to get into profit. Based on this, I'd go for the system which was tested over the larger number of bets. I reckon (going only on the results of my experiment, and without being able to give a mathematically sound explanation why) the system that is tested over 2000 bets is *far* more likely to be in profit as a result of it actually exploiting an inefficiency in the market than the system tested over 400 bets.

[zero]

Its all down to the person me myself i would side with a system which has a bigger number of bets passed through it even if it gives 5% return as for me this shows the system gives a steady return over a long period,but everybody has their view.

[Jonne]

Brilliant post PhantomDude!