Home Forums Main Forum Game Theory and Trading

  • Author
    Posts
  • Curve Advisor
    Keymaster
    Post count: 612
    #1204 |

    Game theory comes up in many different areas of trading. One of the reasons I enjoy trading is that I enjoyed playing all types of games when I was younger (board, card, computer, etc). I wanted to discuss in this thread what I thought were some of the more interesting game theory applications in trading.

    In poker, there is a concept called “levels,” which referred to how the players at the table (including yourself) think about the game.
    Level 1 – What cards do I have?
    Level 2 – What do my opponents have?
    Level 3 – What do they think I have?
    Level 4 – What do they think I think they have?
    and so on.

    There is something similar in trading. You are going to find people who are fixated on what their particular view / positioning is (Level 1). A slightly smaller number will think about what the markets (opponents) are thinking / pricing in (Level 2). But an even smaller number will think about what others in the market think everyone else is thinking / positioning (Level 3). I’m not saying you need to start thinking at some absurdly high level. But the more different perspectives from which you think about the markets, the better off you will be.

  • Curve Advisor
    Keymaster
    Post count: 612

    Obviously, poker and the markets are different, but the basic concept of understanding your opponents still applies. . Here is a very basic example:

    The consensus on Bloomberg is for ADP to be 200K. I think it should come in closer to 250K. Therefore I should sell ED9 right before the number and take profit soon thereafter.

    This is a line of thinking that goes on all the time in the markets. However, there are several ways this line of thinking can be “enhanced:”

    * think about what is the “consensus”
    * think about how you expect others in the market to react before the number
    * think about how you expect others in the market to react after the number

    I will discuss each in detail.

  • Curve Advisor
    Keymaster
    Post count: 612

    ECONOMISTS determine the “consensus” you see printed.

    TRADERS determine the market prices for various instruments.

    There is a common misperception that the “consensus” is what is priced into the markets. This is false, because the the consensus and market prices are determined by two different groups of people. It is very possible that the two could be aligned, but it will frequently happen where it will not. You regularly see where some number comes in different from “consensus” expectations (or analyst expectations), and you get a move in an unexpected direction. This is not an accident.

    Move away from “Level 1” thinking, and try and anticipate what you would expect others in the marketplace to be thinking. In sports betting, people always talk about which way they expect the casual bettor to go. The more experienced bettors will generally take the other side, after the recreationals move the betting line enough. There are many more sophisticated traders in the financial markets. However, you always need to think about what the rest of the market (i.e. your competitors) are pricing in.

  • Curve Advisor
    Keymaster
    Post count: 612

    Not all economists are created equal.

    Yet, in the consensus, each estimate counts the same. Methodologies for determining forecasts range from advanced econometric models to glorified finger waving. As you read various economics’ pieces, develop an opinion of who is good at forecasting data an who is not. Accuracy may vary from data to data – someone good at payrolls may be terrible at retail sales.

    Determine what the “better” economists are thinking, and compare with the overall consensus. There may be a trading opportunity if there is a large enough disparity. Sometimes, you may get a move in the days leading up to the data release, as more people subscribe to the “better” view. If you find this type of analysis useful, you can develop your own historical analysis of who has been the best for a particular piece of data, over time.

  • me
    Participant
    Post count: 27

    Do you go through the process of creating payoff matrices? Or, do you have any useful frameworks / organisational tools for considering all the various scenarios in your game theory analysis?

    It becomes quite complex, especially once you start considering in the second and third order knock-on effects.

  • Curve Advisor
    Keymaster
    Post count: 612

    I think most people just think about how many basis points something will move up or down in certain scenarios. But it’s good to start thinking about the other dimensions – slope and curvature. How will the slope change in various parts of the curve in various scenarios? How will the curvature change in various parts of the curve in various scenarios?

    It becomes difficult thinking about more than a few scenarios. Typically, what I do is take the baseline for what I think the markets are pricing in. I then look at up to four scenarios:

    1) what I think would happen to the curve on the most likely “bullish” event,
    2) what I think would happen to the curve on the most likely “bearish” event,
    3) what happens if we just roll down the curve, and
    4) in rare instances, I may also look at what happens on a “crisis” move (if I think the tail risk is high enough).

    Obviously, if you have enough of a view on more scenarios to be able to say something about a curve move afterwards and probability weight them, that’s better. But keep in mind you may be attempting to produce a precise output when your inputs are imprecise.

    [to be continued]

  • me
    Participant
    Post count: 27

    Sorry maybe I wasn’t very clear. Perhaps what I was thinking about wasn’t exactly game theory per se. But rather the situation when you have perhaps two events (A & B), whose outcome are uncertain, and with different impacts on asset classes, curve, etc.

    So you’d have four cases:
    1.) Only A
    2.) Only B
    3.) A and B
    4.) Neither A nor B

    Typically, I’d look at a conditional probability type analysis, where you’d say hold one constant event constant. If you assume A happens, what’s the impact if B happened or not? Are there some decisions that are good or not good regardless of the outcome of B. What’s the dispersion of outcomes.

    The ‘priced-in’ analysis adds a layer of sophistication, which perhaps this rather academic approach lacks. But, if traders are considering both events in parallel, I’m not sure if I can tell what is their view of each event independently, to make an assessment of how they would react to each event viz-a-vis the current consensus.

    I suppose I am making an implicit assumption of their positioning in my forecast of direction in A/B. But, I think I am skipping a step here without making it clear.

  • Curve Advisor
    Keymaster
    Post count: 612

    I was speaking about one event, but I see you were asking about multiple events. Looking at multiple events gets much more complicated, because there will be some correlation between the events. Let me think about this some more.

    In my last post I wrote “keep in mind you may be attempting to produce a precise output when your inputs are imprecise.” So again, it’s not clear to me if there are benefits to doing a very complicated analysis, when our analysis of just a basic scenario may have a large degree of error. On the margin, there would probably be some value in going through multiple event scenarios, but it’s not clear to me if it would be the best use of time. An analogy would be taking inputs rounded to the nearest thousand and spending a lot of time doing a complex calculation to the nearest dollar.

    My first thought right now is that perhaps we can avoid doing the multi-event scenarios to begin with. Rather than dealing with one instrument that is affected by multiple events, perhaps a better approach would be to pick the best instrument that would isolate a particular event that we have a view on. So rather than try and look at the effect of Greece and China on EDs, trade Greek bonds and Chinese equities to express any view you may have.

    However, you may still need to do the multi-event scenario analysis if you wanted to see what kind of effect the rest of your book may have have (to determine trade/portfolio correlations, or stress testing, etc). Most of this type of analysis is typically based on historicals, but we are in some unusual times, so looking at the past year of data is not really going to be particularly meaningful.

    Again, I’ll think about this some more and post later if I have any additional comments.

  • me
    Participant
    Post count: 27

    Fair point.

  • Curve Advisor
    Keymaster
    Post count: 612

    Just to continue my thoughts on looking at scenarios for one event:

    When looking at the above scenarios (and in particular scenarios 1 and 2), you may be able to isolate a part of the curve (direction, slope and/or curvature) that performs well in multiple scenarios. For example, you may see a particular fly perform well in a bearish scenario, and think that it would be well-supported in a bullish scenario. Those are the types of non-linear payouts you are looking for. What would NOT be useful is a structure that would go up on a “bullish” event and down on a “bearish” event, in roughly proportional magnitudes. A structure could still be worth a look if it goes up a lot on a rally, but goes down small on a selloff. I call this property “non-linearity.”

    The trade that stands out from looking at the scenarios does not always have to be a fly. It also happens with spreads, and in rare instances, it can even happen with certain contracts. An example of a spread performing well would be if you bought ED5-ED9 spread because on a rally, you thought ED5 (or the whites) would lead, and on a selloff the long end of the curve would lead. An example of a single contract could occur from a similar scenario – if you liked being short the blues because on a rally you thought he curve would steepen with the reds doing most of the moving (so you would not lose much), and on a selloff the long end would lead.

    There are many instances where a bullish outcome and a bearish outcome could be led by different parts of the curve. So don’t just think in terms of direction – think in multiple dimensions and also consider the shape of the move.

  • Curve Advisor
    Keymaster
    Post count: 612

    Here’s another example of game theory involved in trading. The previous examples were regarding the “game” you play vs the rest of the market in a negative sum game (a zero sum game with costs). You should always be looking for “edges” wherever you can get it and optimize your trades for those edges vs the market.

    However, you also play slightly different “game” with the firm you work for. I don’t mean this in a disrespectful or dishonorable way – but you are given parameters around which you are judged with your company (or with your investors), and you need to optimize your performance vs those parameters. At some firms, it may be enough just to make money. But typically, you will have capital constraints, drawdown constraints, P&L volatility constraints, Value at Risk constraints, maximum trade size constraints, monthly/quarterly/annual requirements, Sharpe ratio requirements, etc. So you need to optimize your P&L with respect to all of the relevant factors. The most important factor is usually your total P&L, but depending on your situation, this may not be the most important metric at your firm.

  • Curve Advisor
    Keymaster
    Post count: 612

    One of the things you should always do when trading – especially in a zero sum game like futures trading – is to get a sense as to what people in the markets are thinking. You can think about where you think the “value” of a trade is in a vacuum, but the potential move of a structure is going to be a function of what everyone else in the market is thinking. Don’t just look at the tree (your trade)… see the entire forest. What does the market’s positioning look like on that part of the curve? For example. if the consensus view is already for a trade to move in a particular direction, the potential move of a trade is going to be less in that direction because of all the other people who may want to take profit. Conversely. on a stop-out, the move can be even more exaggerated. I’m not saying that there won’t be times when the consensus trade isn’t going to be right. But what you should do is know what the prevalent views and positions in the market are. Always remember that trading futures is a zero-sum game, and that you just one participant in a mutli-player game.

    If you work in an office, I suppose you can find out where others think there is value on the curve. You can also read the research pieces from the various banks and other outlets to see what they like doing. Some of you may even find some value in a newsletter! You can also listen to various media outlets. Finally, you can get a lot of information just from looking at the curve and looking at the open interest. The information is out there, if you just open your eyes.

  • Curve Advisor
    Keymaster
    Post count: 612

    So what do I like to do to get a pulse of the market? Well, I suppose most people do the “usual” things like reading news, talking to people, and watching Bloomberg TV. I think Bloomberg >>> CNBC. I have no idea why CNBC insists on broadcasting with the background trading noise on. It’s not like anyone uses the trading pit any more. C’mon! I’m an old man who finds background noise confusing! And from what little I’ve seen, CNBC seems a little more equities-focused.

    But my secret is… I’m a big fan of podcasts. Call me Mr. Multitasker, but when I am driving or working out or even taking a nap, I like listening to podcasts. What I listen for are not “good ideas” – I mean good ideas are hard to come by. I just like hearing a variety of ideas… and it doesn’t bother me if they are dumb. If some head of research at a major bank is putting out something idiotic, I want to know about it. I also like getting a balance of opinions so that I can weigh the merits of each case. You add this type of market research with what you can infer from watching the STIR futures markets, and you can get a reasonable feel for what is going on and what is being priced in.

    So what are my favorite podcasts?
    1) Bloomberg Surveillance (daily). It’s a 30-50 minute summary of 3-4 interviews that day. They usually get a good sampling of money managers and academics. The host is a little excitable at times, but it can be informative.
    2) Macro Voices (weekly). It’s a hedge fund manager interviewing people he thinks are interesting (usually other hedge fund managers). The interviews are usually about 30-45 minutes long and the host throws in his thoughts on the markets for 15-30 minutes. A few times an episode, I find myself thinking… “mmm… punters.” But it can be entertaining. I stumbled across it because a former colleague of mine was interviewed on it. Again, you are looking to understand the thought process in the markets. And if you get an occasional good macro idea, then all the better.

    Those are by far my top 2, and they are free. I also listen to other things like Freakonomics. I’ve tried listing to a few others (Albert Lu, Macro Musings, Macro View), but I will only sporadically listen to those. An interesting potential source of information could be realvision.com. It’s the same concept as Macro Voices (but with higher profile guests and video instead of audio). I have not signed up for it though – mostly because there are only so many hours in a day. It’s normally $364 per year, but I they have a $100 off promo (check Twitter) for Valentine’s Day so I will check it out. They have a 7 day free trial, which in trading parlance is like getting a free option.

    Let me know if there are things you like reading/watching/listening to to get a pulse of the market.

You must be logged in to reply to this topic.