What happened to Ed Seykota?

[test_user]

If you design a system for max 40 to 50% drawdowns there is no reason to ever lose much more than that.

Here is some deep wisdom for you:
"keep bets small and then systematically keep reducing risk during equity drawdowns. That way you approach your safe money asymptotically" - Ed Seykota

That's not true. You may lose much more than that if you hit a larger drawdown than your historical backtests show. The quote you provide is about keeping your risk at the same proportion to your equity (that's constant % risk per trade, e.g. 1% per trade). In a drawdown your position size would be reduced automatically as your equity shrinks. That doesn't mean you can't lose more than your backtested max drawdown.

I don't think he is suggesting non-proportional bet sizing (e.g. reducing risk per trade from 1% to 0.5% when in a drawdown). Here's a quote ("deep wisdom") by Ed Seykota for you too:
"In this and in your prior send you suggest non-proportional risk management. This raises problems:
...
After a drawdown, prohibitively small risk allotment, effectively suspends trading altogether."

Getting back on topic: you completely missed my point. Check the performance of trend trading systems starting from 2011 (inclusive). Then check the performance up to 2008 (inclusive). Imagine you build a system and tune your risk parameters at the beginning of 2009 and do a thought experiment of what happens in the following decade. It is not a rocket science.

 

[Stockolio]

If you design a system for max 40 to 50% drawdowns there is no reason to ever lose much more than that.

Here is some deep wisdom for you:
"keep bets small and then systematically keep reducing risk during equity drawdowns. That way you approach your safe money asymptotically" - Ed Seykota

Keep bets small ? Even his wording points to a gambling degenerate. How many good trades with a significant probability of happening in value and also in trending momentum... Aren't many a single human can find at any given time.

Every time he says the word business, swap it for trade

 

[southall]

That's not true. You may lose much more than that if you hit a larger drawdown than your historical backtests show.

When designing systems we all know that future drawdowns can be much worse than anything seen in the past. You should design your systems and risk managment and expectations with that in mind.

And lets say i never want a drawdown more than 20% (or 40% or whatever).
Then i do my bet sizing based on the initial 20% of my account. That is my 'risk capital'. The other 80% is my 'safe money'.

eg. I would apply full kelly or half kelly (or 1/4 kelly) or 5% or 2%, only to the risk capital portion of my account. Which grows over time and will eventually dwarf my safe capital. Im also free to move up the safe money watermark level anytime i want.

 

[test_user]

When designing systems we all know that future drawdowns can be much worse than anything seen in the past. You should design your systems and risk managment and expectations with that in mind.

And lets say i never want a drawdown more than 20% (or 40% or whatever).
Then i do my bet sizing based on the initial 20% of my account. That is my 'risk capital'. The other 80% is my 'safe money'.

eg. I would apply full kelly or half kelly (or 1/4 kelly) or 5% or 2%, only to the risk capital portion of my account. Which grows over time and will eventually dwarf my safe capital. Im also free to move up the safe money watermark level anytime i want.

That's all fine in theory. In reality you don't know the true odds of winning and the ratio of average win/loss. Both are needed to calculate Kelly criterion. You are assuming that these values are fixed and will remain so in the future. Your assumptions are based on the backtest results and those depend highly on the period on which you ran your backtests on (e.g. compare 1970s vs 2010s).

 

[southall]

You are assuming that these values are fixed and will remain so in the future. Your assumptions are based on the backtest results and those depend highly on the period on which you ran your backtests on (e.g. compare 1970s vs 2010s).

No I do not assume this. This is why I use a fraction (eg. 1/4 or less) of Kelly on just part of the account i am prepared to risk (eg 25% of the account).
You got to be pretty insane or pretty naive to use full kelly on your full account value. Anyone doing that is asking to blow up.