Contradiction between BJ and Trading?

[blackjack avenger]

Originally Posted by moo321

You can run your strategy through historical situations. For example, how would this have performed over the last 70 years. It should give you a decent picture of how it can be expected to perform.

or
Past Performance May Not Be Indicitave of Future Results. Are we not in a world that is changing? I don't think the markets or any trading strategy saw 9/11.

[MangoJ]

A much more serious problem when testing a strategy against historical data is: Strategies often contain a couple of parameters (i.e. some thresholds), and the usual approach is to adjust those parameters with the help of historical data to maximize performance on this training data.

Now if it comes to testing your strategy, you have to be careful to use statistically independent data for your test. (If you would use the very same training data for your test, of course you would maximize performance - but in effect you already know the future of your test data, as you are trained on that whole data set.)

Now comes the problem, we only have one history. So use the historic data for either training or testing. Using it for testing is a great idea, if your strategy is based on fundamental analytical stuff (like evaluating a blackjack hand).
Using your historical data for training is okay, but then you need to think of a different method of testing.
Dividing the historical data in two separated parts, one for training and one for testing, might be the best of all worst ideas. But then the training data is very old compared to your prospected use of that strategy.

It is a nice little mind trick to fool yourself into thinking - just because your strategy performs well on historical data, it will perform well in the future. It strongly depends on what your strategy decisions are based of.

[Thunder]

Originally Posted by blackjack avenger

or
Past Performance May Not Be Indicitave of Future Results. Are we not in a world that is changing? I don't think the markets or any trading strategy saw 9/11.

BJA,
While the markets didn't see 9/11 happening, it was already in a steady decline at that point. 9/11 just may have accelerated the decline but I don't think it was the cause of the recession then.

[Thunder]

Originally Posted by MangoJ

A much more serious problem when testing a strategy against historical data is: Strategies often contain a couple of parameters (i.e. some thresholds), and the usual approach is to adjust those parameters with the help of historical data to maximize performance on this training data.

Now if it comes to testing your strategy, you have to be careful to use statistically independent data for your test. (If you would use the very same training data for your test, of course you would maximize performance - but in effect you already know the future of your test data, as you are trained on that whole data set.)

Now comes the problem, we only have one history. So use the historic data for either training or testing. Using it for testing is a great idea, if your strategy is based on fundamental analytical stuff (like evaluating a blackjack hand).
Using your historical data for training is okay, but then you need to think of a different method of testing.
Dividing the historical data in two separated parts, one for training and one for testing, might be the best of all worst ideas. But then the training data is very old compared to your prospected use of that strategy.

It is a nice little mind trick to fool yourself into thinking - just because your strategy performs well on historical data, it will perform well in the future. It strongly depends on what your strategy decisions are based of.

Personally, I'm of the belief that if a strategy historically has performed very well given a huge sample size, you'd be wise to keep using that strategy until proven that it does not work. There are no guarantees really in life, but all we can do is make decisions based on the most prudent data we have available.

[moo321]

Originally Posted by Thunder

Personally, I'm of the belief that if a strategy historically has performed very well given a huge sample size, you'd be wise to keep using that strategy until proven that it does not work. There are no guarantees really in life, but all we can do is make decisions based on the most prudent data we have available.

Exactly. We have massive amounts of data going back about 80 years, and thin data going back hundreds of years for some companies (East India Company), that buying value stocks outperforms the market.

It's not overly complicated, it's just that there's so much voodoo investment advice out there that everyone is convinced there's no way to win. I've been beating the index for several years now, just by buying undervalued stocks.

Does playing good cards ALWAYS work in poker? No, but if you play only good cards and bet them, and everyone else plays everything and calls, you'll win in the long run.

[Percy]

Thank you for everyone's contribution to this thread. It's been interesting reading.

For a casino game analogy on analysing historical data for developing strategies I recommend people read Chapter 75 of JG's book.

[blackchipjim]

Sorry for getting into this thread a little late but better than never. So buying stocks and selling them by historical data is pretty much the norm for people in it for the long haul? I pretty much rely on my broker's advice to buy and sell because they are in tune with the market daily. Sound advice with risk taken into account is what I'm comfortable with. I would be able to call in a big player at a blackjack table because I know what's going on at that given time. Now don't ask me about the market because I still think that we are overdue for a big correction but it's still not happpening.

[blackjack avenger]

When you pay for info it needs to add value? I think the benchmark is the brokers advice needs to beat the index that your stocks are in? Otherwise, you could have invested in the index and saved on the broker expense?


or throw darts at the stocks page of the newspaper

[moo321]

Originally Posted by blackchipjim

Sorry for getting into this thread a little late but better than never. So buying stocks and selling them by historical data is pretty much the norm for people in it for the long haul? I pretty much rely on my broker's advice to buy and sell because they are in tune with the market daily. Sound advice with risk taken into account is what I'm comfortable with. I would be able to call in a big player at a blackjack table because I know what's going on at that given time. Now don't ask me about the market because I still think that we are overdue for a big correction but it's still not happpening.

No, we don't buy based on historical data. We analyze a strategy with historical data, and invest with current data.

Historical data could be voodoo "chart analysis".

[blackjack avenger]

With the Market we can run anything we want through historical data, but in the end we are hoping our theories work in the world of today. This is not the world of the 50s, 60s, 70s, 80s, 90s or even last year!

With bj, sims tell us what we can expect from one hand through a career including variance, of course this does not include human frailities plus real world obstacles.

There is far more certainty with bj.


good cards