Tens of millions, I don't remember the exact figure (although if we're being pendantic, it was an error by someone I was supervising. I don't make mistakes ) At the fund where I used to work we had a policy that any errors we made would be for the customers benefit, but the management company would take any loss. Sounds reasonable right? But bear in mind an error could include something like miskeying a parameter in the third decimal place. Because of the size of the fund if that was un-noticed for a few weeks it could easily accumulate into a huge $ difference. Whether that was + or - was just a roll of the dice. Basically the clients had a free straddle on a tiny but not irrelevant fraction of our fund performance. However, these parameters were themselves effectively estimates from a backtest optimisation process. Something like, should the risk weight on such and such be 2.001% or 2.002%? Anyone who knows anything about optimisation knows that the 95% confidence range for that figure is probably somewhere between 0.5% and 4.5%. 0.001% is just a tiny grain of sand on top of a huge steaming Bayesian pile of poo. Really we should only have been ponying up for something that was more than 10% off. If I am ever stupid enough to launch my own fund (and please shoot me if I look like I'm trying), that will be written into the docs. But at the time we paid, and people got fired for this sort of thing (though never for the first offense - they weren't monsters). Mind you I probably made them a few yards on the back of models that just happened to have a long bias during a huge bull market in bonds, but I'm not sure that doing my job (just not as well as I should have done) counts as an error. In the low six figure range, back at the start of my career when I was a lowly junior trader on an options desk, I have a very clear memory of my biggest mistake and largest accidental win. The mistake was when I was settling expired swaptions on about my third day; basically market convention was the trades were supposed to be physically settled into swaps but if mutually agreed you would cash settle them against the ISDA fix (those gentlemanly days are gone, since the funding environment has changed and the value of the optionality has increased substantially). I made an error cash settling to the tune of well over a hundred grand. I was shitting myself - it was about 4 times my annual salary. My MD (who looked like a shaven head bouncer, despite a Phd in computer science), picked up the phone to the counterparty and told him that he was a very naughty boy for taking advantage of a neophyte trader, and that if he didn't reverse it then nobody would ever trade with him again, as he clearly wasn't a gentleman. He reversed it, and I breathed again for the first time in an hour. My biggest accidental win was when I was trying to guess the correct premium to add to an american swaption over the bermudan (we couldn't price the American properly with our crappy software, which was obviously more valuable than the bermudan but we didn't know by how much). My guess was way too high (good for us) but the counterparty didn't know what they were doing and took the price. They were an internal counterparty (the equity structured products desk), and they could only value the trade by taking our valuation and putting a minus sign in front of it. My counterparty shat himself. Now if this was a fairy story the MD of the equities desk would have called my MD and asked him to reduce the mark to a less rapacious level or nobody would ever trade with them... blah... blah... blah, and being gentlemen they would have agreed. In fact what happened is that my MD told the equities desk they were idiots, deserved everything that was coming to them, and could fuck off if they thought they were getting a penny out of us. The difference is that the equities desk had to deal with us, no matter how mean we were to them. (I have a lot more of these stories, so maybe I should hold them back for another book) More recently, back in the real world, I've probably had a few errors in the same range as you, mostly due to crappily written software (I'd fire my CTO but he works for free)... GAT
In an identical situation at an adjacent firm, I price-checked the internal desk against the IDB and ended up trading with the with the street. You can imagine the type of anger that creates - how dares he not trade with us at inflated prices and instead going outside!
Yes, we were supposed to hedge our delta with the internal swaps desk. Every now and then we'd trade outside, just to keep them honest... GAT
Thanks for sharing Rob! Very entertaining read. Are you and your CTO are same person? Looking forward to hear more "retail" stories if anyone is willing to share. Val
Hello, I am a new comer here and so far enjoying your posts! Can you please elaborate on what do you mean by "execution" and "account size" edges? Thank you. N
HI Nathan. While nothing about trading is easy, the smaller your account size the easier it is to make higher % return on your money. A simple example will be entering the market at extreme without impacting the price much. If you need to open or close a 1mil$ position you can only do so fast and without suffering too much slippage in a highly liquid market. Which limits your choices. My account is small enough be able to get in and out very fast and still being able to trade about half of individual US stocks. It's also a matter of competition. It's like being a lightweight boxing champion. Are you a champion? Yes. Did you compete with Mike Tyson - no. Btw, good tips on picking a good fighting opponents can be found in this video. If trading would be a science, everyone's PL would have been normalized relative to their account size and imposed restrictions on what they can and can't trade. Execution edge is a combination of Automation and order execution tricks. In my case - anywhere between hundreds to thousands of orders are handled in a single day. It is very feasible to make money trading sinplier version of my systems, but with automation I can do more of what works, faster and with less errors than a human would. That's what I call execution edge. Otherwise I trade same order types as everyone else, with same retail broker. I'm actually thinking of doing a zoom session over a weekend, to talk a bit about the process I use and what steps are automated followed by Q&A. I'll keep you posted. Maybe I'll do it this Sunday. Val