HFT, and is backtesting helpful or harmful?

[Redneck]

What do you mean by "all signals are signals", Redneck?

Bring up any historical chart..., (prior to right now) - the signals are there (they are formed) - the outcome even known (which is what makes/ validates the signals - as signals)

I don't live in that world

==============

Realize - I am not saying back testing is useless for everyone - I'm no where near qualified to say that....

I am saying;

I've never found it beneficial..., hell.., I couldn't even figure out how to make it beneficial - for me

RN

 

[xelite777]

most liquid market on earth comes with plenty of variables

What variables, Bob?

where in stuff that i use to trade- very few. which one ,logically, should be easier to trade profitably for trader of every level?

Still not sure what you mean by variables. Anyway, avoid illiquid markets and you will be fine, that's all I am saying.

why do you even care about all this stuff like HFT, backtesting etc?

Quite frankly I don't give a rat's ass about HFT, my FX systems are still profitable and absolutely nothing has changed. Zero, nada.

Backtesting I do care a lot, as I am always backesting new trading ideas. A real trader must always learn and experiment with new strategies (even the seemingly weird ones) and.... boldly go where no man has gone before!

 

[Dustin]

1) Does HFT really affect daytraders PnL?

For me, yes it does every day. I use every trick in the book to find liquidity in the names I trade. This can include midpoints, darks, hidden, pegs, or discretionary orders. Even with all of this my orders get gamed day in, day out. Maybe it costs $100/day, maybe $1000/day...it's impossible to say.

Just one of the blatant examples today when I was trading CSLT which averages 1.3m shares a day. I've been trading it from the long side, building positions and selling on spikes for days. Today there were 1100 shares offered with a 5c spread. Nothing was happening in the stock, but I wanted those 1k shares so I tried to take with my smart route. It took the cheapest first (100 shares), then tried to get the rest. The 1000 were taken out from under me and the stock spiked 20c instantly. Sure, someone could have put in an order 20ms before me...but it's highly unlikely that that is the cause every time. This cost adds up big time over the course of months/years.

Newer traders don't remember the days when posted size could actually be taken, so maybe they think it doesn't affect them...but we all pay a daily tax to hft whether we know it or not.

 

[xelite777]

I've never found it beneficial..., hell.., I couldn't even figure out how to make it beneficial - for me

RN

Ok, got you.

But then, how do you know that your system WILL make you money, in the long run?

What makes you believe that your trading strategy HAS indeed a real predictive power, if you don't test it first with enough historical (or live) data, assuming the trading rules of your system are clearly defined?

Just curious.

Uh-uh, I have an unconfirmed long signal on the AUD/CAD, got to watch this now

 

[Redneck]

Ok, got you.

But then, how do you know that your system WILL make you money, in the long run?

What makes you believe that your trading strategy HAS indeed a real predictive power, if you don't test it first with enough historical (or live) data, assuming the trading rules of your system are clearly defined?

Just curious.

My system (me) has been making money trading for the last 15 or so years

My system (me) has absolutely no/ none / zero/ nada/ zilch/ not one damn bit of - predictive anything

RN

 

[Dustin]

2) Is backtesting helpful for a new (unprofitable) daytrader?

I think a lot of new guys put the cart in front of the horse. The idea of coming up with a magical system that the computer can test and trade for you makes new guys see dollar signs. My world isn't that big, but I don't know anyone who has succeeded by starting with system design before learning to trade first. I suggest everyone learns to trade with tiny size...there is no substitute for experience.

Here's one big problem I have with back testing, and admittedly it applies to my style of discretionary trading. I could look through my blotter today (see attached) and find very few similarities in the charts and my trades. When you trade volatility there are probably 30-40 different strategies you can use to make $. I traded 59 different stocks, with 176k shares, hundreds of fills, dozens of ideas. Not to mention, every day is different! It would literally be impossible for me to backtest my trading.

 

[xelite777]

Here's one big problem I have with back testing, and admittedly it applies to my style of discretionary trading.

That's the problem with discretionary systems, they are extremely hard or even impossible to backtest! Not only that but the risk of curve-fitting them is high.

Why would anybody want to trade these almost-impossible-to-test types of systems instead of sticking with simple but powerful mechanical systems with clearly defined rules is beyond me.

I did not say they are unprofitable, but why not start with easy to test trading systems first?

 

[Craig66]

Dustin, for me your posts kind of contradict stuff that other people who I respect have said. It kind of sounds like you scan for stuff that's on the move then wing it (or at least reconcile situations repeating). Obviously it's working for you. But then you have Eric & Corey who seem highly structured, Eric more than Corey (all this of course is guessed from their posts, so I could be way off base). So it seems successful traders span the spectrum from no testing to all testing. The trick is (I guess), if you are a tester, is isolating situations which are worth testing. This of course, does not come from back-testing random stuff, however, it can come from learning about market structures and relationships, participants or from observation.

 

[Dustin]

That's the problem with discretionary systems, they are extremely hard or even impossible to backtest! Not only that but the danger of curve-fitting is high.

Why would anybody want to trade these almost-impossible-to-test types of systems instead of sticking with simple mechanical systems with clearly defined rules is beyond me.

I did not say they are unprofitable, but why not start with easy to test trading systems first?

I suppose I would just ask you to show me someone making a living from "easy to test trading systems".

 

[Dustin]

So it seems successful traders span the spectrum from no testing to all testing.

Yes, but they typically begin as traders. That's where the ideas to test come from. Just imo...