Wednesday 4 May 2011

Measuring Weakness


The consensus among traders is that a sub-category under the heading of "Discipline" is that of keeping accurate records, but it is arguable this may be more of a hindrance than a help at times. The previously discussed Dr. J thread had a comment to this effect

The worst thing about keeping note of good betting runs is that we are setting up artificial hurdles for ourselves. We are creating a difficulty that should not exist.
As an illustration, as April was winding down, I was very much aware of how close to ridding my spreadsheet of the one red month I was. While I tried not to let this affect my betting, the seed was planted in the back of my mind, and undoubtedly impacted my decision making process. I had the short-term goal of making a large enough profit in April to cover previous losses, which meant that I was more selective than I would normally be.

While it could be argued that this target was meaningless and artificial, to me it was important. Passing any milestone is really meaningless, but going from say £980 to £1,001 is probably more exciting to most people than going from £1,001 to £1,023, even if the latter is the bigger win.

How would I have bet if April had been close to a record month, and I needed 'just' another £300 on the 30th to beat my personal best? Would my betting have been a little looser than normal in an attempt to beat that artificial goal? While we should always play the same way, I doubt that too many of us have completely mastered this skill.

One suggestion on the Betfair Forum from a few years ago, was to trade without your balance or P&L figure showing on the screen - the idea being of course that your decisions would be based on the current situation alone, uninfluenced by factors such as those I have mentioned above. Easier said than done, but the idea does have some merit.

There was another comment that was mildly amusing, given that the thread starter has made close to £200k in seven years. After reporting that recent results had been poor, someone posted this gem
...but I now see a real weakness - and maybe you have to convince that your edge has been tested sufficiently over enough events
OK, poor English, but you get the idea. The number of events that the commenter felt was required to prove an edge exists? Here's his answer
I know we are digressing from the main theme of the OP [Original Poster], but to me the 'edge' has to be proven over at least 100,000 events before you can claim it being successful.
100,000! Talk about an artificial number. 99,000 won't do, but roll that odometer over to six digits, and now we're on to something real!

One reply, which I thought was particularly brilliant, was:
Any edge in the markets only exists because there is an inefficiency, and will not last forever. The edge may be 'proven' one day, and 'proven' to be gone the next day. However, in markets, as in plate tectonics, as one edge closes, another one may open. 100,000 trials in the context of sports betting is nuts. At 20 bets a day, it would be close to 14 years before you had that number of bets, and to say that the markets might have changed in that time would be putting it mildly. And if you have a system that produces 20 value bets a day, you should be living in the Cayman Islands long before 14 years has passed by applying Kelly.
At the rate my Strong Draws are showing up, it'll be 645 years before I know if I have an edge or not. The Methuselah Foundation had better get moving.

4 comments:

mouldhouse said...

100k iterations, what a joke.

It amazes me with regularity that people don't understand what's statistically necessary to "prove" or otherwise a "system" or method.

100,000 bets on horses that are all priced 200/1 or above might not be significant to "prove" anything. 100 bets at even money with 70 winners and 30 losers will provide confidence to a certain level that there is a winning methodology at work.

It does correctly identify though that just because there was a historical edge, all these markets and edges are dynamic, and past performance is no guarantee of future success!

Imagine paper trading 100k bets before placing one. You'd need to be a millionaire to afford the paper and pens :)

BotTrader said...
This comment has been removed by the author.
BotTrader said...

As a trader I see no point whatsoever in keeping records maybe they're helpful for all the system/stats bettors. I've done it in the past, thought it was the thing to do but for someone taking ticks here and there I can't see what use it'd be for me to know how well I did in August last year.

I'd recommend trading with the balance off as continually checking the pnl can be counterproductive by either making people take unneccessary risks when they're down or being overcautious to protect profits when up. If I take a big hit on a market I'll check the 30 day or 7 day pnl on the betting history and that's usually enough to convince me I'm doing something right so as to not overreact.

As for birch's idea of needing 100000 bets to prove an edge it's quite obvious he doesn't have an edge. Anyone with an edge knows exactly how it works and when it occurs, you certainly don't need 100000 bets to prove it to yourself or anyone else before using it.

inplay said...

Hi. I'm a relatively new trader. I mainly trade pre-race horse markets. In your opinion, how many bets/trades would be a good marker to see if you do have an edge? Or is it a case of if you are in profit in 3/6/12 months then you have an edge. I realise that you could lose your edge over time, but just wanted your opinion really. Also, as far as keeping records go, to an extent it is helpful in my opinion. Not just in the sense that you know what you have won/lost but you can see which bits you can improve upon, what markets are not that profitable over time or as you have mentioned before in previous posts, which months are not that kind to you.