Interesting Article.
- Thread starter Craig66
- Start date
The problem comes in with order handling algo, at order matching end. If your able to buy the bid and continue offering inline with the signal. Or selling at the offer, and continue bidding inline with the signal, till risk management algo takes you out.
Easier thing to do is partner with a pit trader, so entries are always made with the 'edge' and have phone relay with hand signals into pit. Even though pit traders have the edge, and extra layer of edge is for entries with the edge to correlate with temporal directions.
IOM seats.. 60-70K at CME now.
I think during my best months I might have been making (pre-commission) $4-$5 per round-trip russell trade. How did you calculate the 1 tick?
In the article I state that my program did not use length of time in a trade as an input for any calculations. Also, it made no distinction between opening and closing positions. It had no concern with "closing quickly enough" - unless it was the last 30 seconds of the day and it needed to exit a position.
My algorithm had no directional bias - long and short trades were calculated in the exact same way. And since, the market basically works pretty much the same at a small scale (as far as going up and down) the result was that there was an equal proportion of long/short trades.
I guess what people need to realize is that I was ONLY trying to get good prices. If I always got good prices nothing else would matter because in the end I would average out to making money.
No, I think a trader friend had a leased seat at some point but it never made sense for me I believe. But yeah, our trading group had searched around for the best commission rates based on our style of trading and how much volume we did. (I was the only one that created the automated program though)
I don't know, my experience with MM / microstructure is that markets can trend during the day which means you could have a whole day where the price has momentum pretty much in one direction -- and whatever price action rules you came up with will very likely bias your trades in one direction or the other. It would be too much of a coincidence if you were close to 50/50 day after day after day when the markets can go up 1-3% one day, and go down 1-3% another day... And in those types of trend days, I would expect your PL to fluctuate wildly. Still doesn't sound quite right. What's strange is your PL only fluctuates to the positive side...
Now I can guess that you will argue that at the micro scale it all balances out, but that is not really the case. Often you will see a series of large orders that take out one side of the quote repeatedly for several minutes. This kind of thing happens all the time. And there is really no way to anticipate marketable order flow like that: the reason the price jumps at those high volume bars is precisely because there were not enough orders on the book to keep the price stable so you can't really say the order book anticipates all of those price moves.
Now when I asked about the length of a trade, of course, I wasn't asking whether you imposed a time limit -- I was asking what the round-trip trade duration ended up to be with the limit orders. This will have huge consequences because if the trend negates the supposed spread you are earning, then your PL will suffer. Even if it averages out, it might take days or weeks, and your PL volatility will be huge, and your sharpe will take a big hit.
It still seems bizarre that you would give up this money making machine without even bothering to find out your own trading stats (win %, average winning trade, sharpe, etc) which I would think anyone would be curious if they wanted to fix their algo... Also, why not also think about hedging?
Well, it may be interesting to calculate those stats but I'm not sure knowledge of them could help me improve my program.
Wouldn't hedging be something you do if you want to lower risk? My trading was already extremely low risk so that wasn't an issue.
If I did want to stay in trading I would have either A) Tried to find another independent algo trader to join forces with B) Alter my system such that some of my friends who were trading manually could trade through it and thereby get better prices (and hopefully we'd both make more money) or C) Try to get hired by one of the HFT trading shops.
With regard to the microstructure - in the case that the market is trending up all day due to a lot of buyers - it's possible I did less well on those days but I wouldn't be so sure. There are still some people hitting the bid. Plus, my program did cross the spread sometimes. I know when my friends and I were trading manually we HATED those days where there were never any corrections but I think my program did just fine. I can't really argue any further on that point.