Cobra's Market View Private Messaging System shows:

I cannot emphasize enough about perhaps the most important thing about the trading – THE POSITION SIZING. No matter how accurate your trading system is, as long as it’s not 100% accurate, if you don’t **follow exactly the same position sizing rule for every of your trade**, one single lose would make you lose everything. Say, you have a 90% winning system, you risked $100 for 9 trades and each trade won $100 for you, so you decided on the 10th trade, you’d risk $900 which are all you have won in the past 9 trades. Unfortunately the 10th trade was a losing trade and as the results you’ve lost lost everything you won in the past on a single trade despite your winning rate is 90%.

The very basic concept about position sizing below is quoted from http://tradermike.net/2005/07/position_sizing:

Dr. Van K. Tharp did an experiment which shows the importance position sizing. In his book "Trade Your Way to Financial Freedom" Van gives the results of his testing of four different position sizing models. He tested the models on the same trading system, so the only variable was the position sizing. The simulations were run with an initial equity of $1,000,000 and took 595 trades over a 5.5 year period. The models produced drastically different results:

- The worst was the baseline model which just bought 100 shares of stock whenever a signal was given. That model returned $32,567 or 0.58% annualized.
- Fixed-amount model: This method traded 100 shares per $100,000 in equity. It returned $237,457 or 5.75% annualized.
- Equal leverage model: Each position in this model was 3% of the account equity. So at the start of the trial each position was $30,000. This method returned $231,121.
Percent risk model: According to this model positions were sized such that the initial risk exposure was 1% of the account equity. So with $1,000,000 equity the initial risk would be $10,000. So if the initial stop on a trade was $1 the system would trade 10,000 shares. For an initial stop of 50 cents the system would trade 20,000 shares, etc. This model returned $1,840,493 or 20.92% annualized.Percent Volatility model: Positions were sized based on each stock's volatility -- the more volatile the stock the fewer shares are traded. For this trial positions were pegged at 0.5% volatility (initially $5,000 per position) -- so if a stock's average true range was $5 the system would trade 1,000 shares. This model returned $2,109,266 or 22.93% annualized.

As you can see the Percent risk mode and Percent Volatility model are the 2 best models.

And above that, I use the original Turtle’s 2% stop loss rule: Never risk more than 2% of your total capital for each trade.

So combine them together, say, you have $100,000 account, then for each trade, your max stop loss amount is 100,000 * 2% = 2,000. If your stop loss point is $2 below your entry point then the max shares of stock you can buy is 2,000 / 2 = 1,000 shares.

Another very important rule is: NEVER EVER AVERAGE DOWN. If you follow the position sizing rule above then this rule should be implied automatically – you’ve already risked 2% of your total capital for a stock so NO MORE shares should be added to the losing position ever!