The Gateway to Algorithmic and Automated Trading

Spreading out

Published in Automated Trader Magazine Issue 15 Q4 2009

One thing you can say about Bill Braman is that he is a survivor. He’s been trading since 1987, moving from open outcry right through to today’s technologies. How does he do it? Andy Webb caught up with him in Chicago, and asked him.

One thing you can say about Bill Braman is that he is a survivor. He started out as an open-outcry pit trader in 1987, trading spreads under the mentorship of Charles DiFrancesca in Chicago. When trading went from the pit to electronic, Braman did the same, moving to a screen at Goldenberg Hehmeyer. The deal there was that he would help their people with trading and they would help him go electronic.

After reading a copy of the first-ever issue of Automated Trader, Braman approached CQG to ask whether they might be willing to help him make the transition from manual to automated electronic spread trading. And not long after that, CQG approached Braman to ask whether he might be willing to help them develop an automated spreading tool …

Survival is a relationship business. But to survive as a spread trader from 1987 to 2009 takes more than just good contacts and enduring friendships. Our founder, Andy Webb, caught up with Bill Braman in Chicago to discuss spread trading, the development of the CQG Spreader (and other tools/techniques past, present and let's hope they're coming soon), long-term survival in an ever-changing but always-complex market, programming and basketball.

Bill Braman

Andy Webb: Our readers will obviously want to hear about the CQG Spreader, but perhaps we could start with some background on you first. What made you start spread trading? Take it from the top.

Bill Braman: When Charles DiFrancesca started out, a few years before I did, he realised by watching the spreaders that if he traded the calendar spreads on the back months he could immediately start trading bigger size while getting better edges. He always told me that the best way to trade was to be a spreader. At his suggestion I started trading in the ten-year pit, which was then much smaller than the bond pit, and I would just leg the NOB (10yr note v T-Bond) and FYT (5yr note v 10yr note) spreads. I got used to trading spreads; that way you could make a big market, the brokers would trade with you, but since you're laying it off you wouldn't be taking the kind of risk that size generally forces you to accept.

I went to Northwestern for my graduate degree and took some stats classes and on the back of that came up with the concept of standard deviation bands at about the same time John Bollinger came up with the same idea. I wrote an Excel spreadsheet that let me figure out where the NOB and FYT spreads should be trading. Whenever it got out of whack in the pit I would step in and do a few hundred lots.

All of a sudden I'm this new guy in the pit and everyone is looking at me to see why I was trading so much volume and making money. The mean reversion was gorgeous. The busier it was, the faster the moves, the more likely it was to snap back into place.

Andy Webb: And then?

Bill Braman: When the pit looked like it was moving to the screen I switched to be a prop trader at Goldenberg Hehmeyer for two years, to learn how to do it electronically. I kept up the spreading though. That went pretty well until about 2004/2005, when the market started getting very slow and the bid/ask spread very tight, and I was continually getting hung on the legs of my spreads.

Bill during his days in the pit (with a colleague kindly indicating that Bill has just gone long 2 lots of the NOB)

Andy Webb: What were you using for technology then?

Bill Braman: Eccoware and Trading Technologies (TT), but I was also in close communication with some guys at CQG about some of the technology they had coming down the pipe. I started experimenting with that and after a while they asked if I would help them build an auto spreader. I had loads of ideas I wanted to share so I jumped at the chance, which is how I got involved with CQG.

Andy Webb: So you always traded spreads then or did you do some outrights too?

Bill Braman: Yes. I have always managed to give money back by trading outrights, but I've always made money being a spreader. For me, spreading is where it's at, so I wanted to participate in building the best spreader possible when I teamed up with CQG. I told them that it takes too long to leg the stuff remotely, we need to put a server right next to the Globex server so I can leg my NOBs and FYTs. They did that and now they have the fastest thing known, as far as I'm aware, and it's spectacular what it can do in terms of speed and execution. Second-best thing about it, after the speed, is the missed-order management. If you miss a leg and get hung, we have managed to automate every possible scenario that you might want to program in to deal with that.

Andy Webb: Your ideas. Would you say that the way you are trading has changed much in terms of the basic principles you've been using, as regards mean reversion? Have you added some extra statistical modelling on top, or has it just been a case of staying ahead of the pack in terms of speed, and evolving technologically?

Bill Braman: Let me put it this way. When I was using other software a few years ago it was a slow market. You just wanted to buy it at one and sell at one and a half. Or sell at one and a half and buy back at one. But you couldn't get the trades executed. You would be buying at one and get hit on a leg and then miss and buy at one and a half and vice-versa. Now, the thing moves around a bit more. I still have to use my brain to decide where I want to buy and sell it, based on different models, but the fact that I'm only trying to make one or two ticks on a trade and I never miss an execution, that's a huge deal. I still look at the old-school factors and make trading
decisions based on trend, but I think having good execution is what is key to me.

Andy Webb: So the spreader has taken away all of your execution-management headaches, so you can focus on the strategy.

Bill Braman: Right. 100% true.

Andy Webb: That must feel good.

Bill Braman: Sometimes it's hard to scale; because of the algorithms involved, you can't get big pieces of it. In some cases the speed now is important because of the algorithms. If you're legging grain spreads, it's different now because you have to go up against the Globex Implied Spreader and no one can beat that.

Bill Braman

Andy Webb: Your inputs for your trading are obviously the price series of the things you're actually trading, but are you also looking at Eurodollar and other things?

Bill Braman: Yes. You definitely look for clues as to what's going on and which way the market is going. You look at the big picture and still have to make your own trading decisions. I don't have any secret formula for that. I wish I did; I have tried many times to develop one but to no avail. Talking of automated trading, I have not been able to automate yield-curve trading in a longer time frame; what's going to happen in the next fifteen, twenty minutes, the next hour? There's a lot of feel and experience involved in that. I would love to be able to automate it, but I haven't succeeded as yet.

Andy Webb: Have you managed to automate even the very short-term stuff or do you still have to click the mouse?

Bill Braman: I use a lot of tools, the CQG Spreader being the major one, but the main thing is to set up your spreaders, a lot of them. If I was trading grain spreads I would set up thirty auto spreaders just to cover the corn. It's very difficult to leg, say, the July/Sept spreads, because the Globex Implied is going to leg that for you. So what you do is, you set up all your auto spreaders to leg July/Sept against July/Dec in order to leg Sept/Dec, and so on. By using these spreaders you collect edge all over the place and it co-mingles and you get net positions and then you're working your way out of that. So yes in that respect it's automated.

Andy Webb: I've been dealing with CQG for about 15 years now, and the direction they are going fits perfectly with you, I would think? They brought the API out, and now (thankfully) they've improved the documentation that comes with it. The other thing is, they have opened the whole thing up and now have started to work in terms of partnerships with firms that use their API. MATLAB really impressed us recently and now you can get the data out of CQG into MATLAB and fire the orders back from MATLAB into CQG. If you were going to do super-high-frequency stuff, it's probably not ideal, but it means you can shoot stuff from MATLAB back into the CQG Spreader as well, which I'm really interested in for synthetic pairs and so on. CQG seem to be going places; you've worked closely with them on the Spreader, what do you think?

Bill Braman: I'd agree. For many years they were a data provider and they just focused on making the charts. All of a sudden in the last few years they've realised that there is a huge market of traders who want to execute trades and have different strategies, but want to customise them for themselves. CQG are close. They have the data feed with such high penetration of the market that if they can provide the tools, people will come running to take advantage. They've cleaned up the API; there's the thing with MATLAB; they're working with Apama now.

It's also the little things. If I'm trading the NOB and I have Bollinger bands around it on a fifteen-minute bar chart, to sell when it gets up, buy when it goes too low, I can put my Spreader order to follow that line. No matter where it goes, the Spreader will work the order for me. You can do the same thing with fifteen different items. They're allowing you to figure out a strategy and then go away and focus on the next one while it works away.

Andy Webb: I like the extensibility, accepting that they can't do everything and devising a strategy that their customers are not constrained by what they can't yet do. The ability to integrate with MATLAB, for example. But I've gone off the point. This is supposed to be an interview. Let's get back to the questions … you are mostly trading the interest-rate curve … are you still trading the agriculturals as well, or not so much these days?

Bill Braman: I did last year. 2008, when it was really moving, there was a lot of opportunity and that was working out really well. I have a lot of friends who are still trading that, and you can still make money if you're patient, collecting a few ticks at a time. But I've tried to move on to things that are more exciting.

Andy Webb: What's your average timeline for a typical trade, if you have such a thing? Are you carrying positions overnight, or are you in and out within fifteen, twenty minutes on average?

Bill Braman: My yield-curve trading would be typically 3 to 25 minutes timeframe, trying to make, on average on a NOB spread, maybe only one tick on the 10 year side. That's why the execution is so important to gather this small number of ticks. There is no secret formula, just hard work and clean execution.

Andy Webb: Would you say it's the speed of execution or the execution management that's most important. I understand the co-location, having the server close to the matching engine, but it also sounds as though you are pretty pleased with not having to worry about the legging-risk problems you'd normally get.

Bill Braman: I said in the first meeting with CQG: "It has to be fast but when you miss, I need this, I need this, I need this. I want to follow the bid, I want to see how many lots are left. I don't want to work an order unless there are this many, or this is happening, this situation or that."

The thing is, you can't control it all with a mouse click. Your hung orders you can handle, but when automation's there, even if you miss you generally get it on the rebound, or in the worst case maybe lose a fraction of a tick. In terms of automation, as a trader I'm very old-fashioned yet I'm using these new state-of-the-art tools to allow me to trade at the price of my choosing, rather than the price I end up getting.

Andy Webb: Very roughly, would you be able to say what your trade volume is? Not in lots, but your transaction count in an average day?

Bill Braman: Not too many. 500 to 1,000 a day.

Andy Webb: How many contract months are you trading across? And by the way, do you trade the Eurodollar?

Bill Braman: I use it more as an indicator and usually don't trade it anymore. I like to watch the spreads on the Eurodollar and the Board of Trade products as they are a huge indicator of where it's all going. Although the Eurodollar guys tell me they look at the ten-year trade while we're looking at their trade! They are highly correlated. About 9 to 15 months out is a good part of the curve that leads the five and ten year. That's a good indicator.

Andy Webb: You say you're mostly trading between the 5s and 10s. How much of your stuff is intra-month, between contract months in the same instruments?

Bill Braman: There isn't much to trade in the back months for what I'm doing. I'm not here to speculate about what the calendar spreads are going to do. That's hard. The cheapest to deliver is changing from time to time and that's not my game, but I do trade it for about 2 and half weeks every quarter at roll-over time. You can also leg your FYTs and NOBs and spreads using the back months; sometimes there's a nice edge in there. You can actually improve your price on a regular trade by instead of selling Dec selling March and get an extra half-tick on the trade that you can cover later when you're legging out of your spreads with your autospreader on your calendar spread.

Andy Webb: What about time frames? You can have mean reversion between the two in one time frame and a different parameter for that in a different time frame. Are you trading multiple time frames at the same time; say, five-minute chart, mean reversion, fifteen-minute chart, mean reversion etc?

Bill Braman: The things I watch, I kind of just trade my feel. When I go into a trade I already know where I'm going to get out either for a profit or loss. I have parameters in my mind based on volatility models. But I still watch the one-minute, the fifteen-minute and the daily chart. I think people lose sight of the daily chart. The big boys trading long-term positions are involved on a daily basis so you need to know when trading is down over the past week in case there is a bias. The fifteen-minute is a good chart for indicating what's happening over the course of a day and the one-minute I use to pinpoint my execution, where I'm actually getting in and out.

Andy Webb: Have you plugged CQG into Excel to run your volatility models?

Bill Braman: Yes. I have a couple of macros I wrote up about fifteen years ago that I still use to this day. I copied some of their code and made some programs and pulled the data. The macro analyses it and spits out levels in a volatility model so I have an idea of my plan for the next day. I also use CQG plugged into Excel to make my own custom studies and indicators and charts. That's so that I can see data in a different way from the way they offer it, in the way I like to look at it.

Bill Braman

Andy Webb: Bill, I'm looking at your CV here, and frankly, you look like a survivor. A lot of guys came upstairs from the pit to click trade and blew up fairly quickly because they couldn't hack it. Then next time around people tried to move from click-trading in an arcade to automated trading and blew up. But you're still standing. Is that because you work on the basis that you can't assume that tomorrow's going to be like today, so you've got to be innovative and come up with new strategies and insights all the time?

Bill Braman: I would say … I'm an idea man. I'm always trying to think of new ways to do anything better. I play a lot of basketball. Whether it's working on my jump shot by adopting new techniques, or seeing new styles in the way people play. Even thinking about a better way to clean a house. Anything I do, I'll say, how can we do this better? My very first years of trading in the pit, I figured out a little model, and all of a sudden I'm standing there all day holding a sheet and whenever the levels hit, I'm just saying sell. People would ask, how do you know? I would say, I don't know, I just looked for a better way to figure this out in advance. I'm always looking for something new. New ways to look at things, new techniques, ways to combine new and old models and keep on pushing forward.

Andy Webb: Your timing seems to have been pretty good, too.

Bill Braman: Maybe I've been lucky. I saw the pit wasn't going too well, and I went to a friend of mine who was running a prop shop at GH and I said, why don't I come and work for you? I can help your young guys with trading and you can help me with electronics. It was the same as we've gone forward with CQG. I was trying to get into automated trading. I had Automated Trader magazine, the very first issue. It was being handed out at the Board of Trade and I had it on my desk. I said: "We need to get more automated".

Andy Webb: Do you handle your own programming?

Bill Braman: I do 100% of it myself. I'm a self-taught, seat-of-the-pants programmer. My last computer course was in 1981 in high school. I have a thing for languages - French, Spanish - and computer languages I pick it up pretty quickly. I get excited about programming, writing code and seeing efficient programmes. Unfortunately most of my stuff is like a bowl of spaghetti, but it gets the job done.

Andy Webb: Bill, it's been great. Thanks for your time…