I confess I have always found it hard to trade options. This is despite having read some of the "bibles" of options trading, including Lawrence McMillan's Options as a Strategic Investment and Euan Sinclair's Option Trading: Pricing and Volatility Strategies and Techniques. Partly that is because I prefer simple strategies, and options strategies are rarely simple. Partly that is because I was brought up on stocks, but stock options are depressingly illiquid. Most successful options traders that I know of prefer to trade index options instead, an area that I unfortunately have no intuition at all. Papers and books written by options professionals on this topic tend to be dense with equations, and worse, they seldom focus on the practical side of trading.
That's why I am pleased to learn that Larry Connors, whose books I enjoy due to their simplicity of exposition, is presenting his first ever quantitative index options trading seminars. Interested traders can register for his free preview webinars on August 9 and 15 here, or a pre-recorded preview here.
===
Speaking of seminars, readers in Asia may be interested to know that my own workshops on Backtesting and Statistical Arbitrage will be held in Hong Kong on October 2-5. The same workshops will be held in London on November 19-22.
(I enjoy giving those workshops very much, because many of the participants are institutional traders whose knowledge and points of view are very much at the cutting edge. Past participants include quants and traders from, in no particular order, Goldman Sachs, Morgan Stanley, Royal Bank of Scotland, Bank of America, UBS, Societe Generale, Deutsche Bank, BNP Paribas, JP Morgan, Barclays, Citigroup, Blackrock, and various other Asian and European hedge funds, energy companies, banks, and asset managers. I humbly submit that the in-class discussions are sometimes more interesting than my prepared materials.)
===
I wrote some time ago about those FX brokers or ECNs where algo-traders can colocate their trading programs to lower latency for a reasonable price. There are also similar options for futures algo-traders. For e.g. Optimus Trading Group provides a market data service called Rithmic which is colocated at the major futures exchanges, and traders can colocate with Rithmic to reduce latency. Of course, traders can also directly colocate at the new CME data center in Aurora, IL. I suspect, though, that the cost of the latter option will be considerable.
====
Finally, as a quant trader, I nevertheless read macroeconomic analyses occasionally, if only to figure out why some of my strategies suddenly start to fail. One website that provides interesting analysis of the energy markets is oilprice.com. In particularly, this interview with economic commentator Mike Shedlock is unusually detailed and thoughtful.
The Mike Shedlock article seems like typical Ron Paul fan whining. Global trade is not going to collapse whether Obama or Romney is elected president.
ReplyDeleteI'll happily take the other side of his gold trade too.
Hi Ernie,
ReplyDeleteI'm just curious, do you usually issue market or limit orders on your pair strategies? I know there is the option to issue limit orders through Interactive Broker's Combination orders but I'm curious as to how that is carried out as well. The combination order will have to immediately hit the market order for security B if the limit order on security A is taken. It also seems quite complicated to ensure that the exact quantities quoted get filled without running down other bid/ask levels.
Do you usually just use the combination order or program your own orders in using your own logic?
THanks!
Hi Anon,
ReplyDeleteI issue limit order on the one leg (the less liquid side), and hedge with market order on the other.
I don't use combo orders anymore: I use Matlab to control executions.
Ernie
Hi Ernie,
ReplyDeleteFor intraday trading, do you still run linear regression based on end of day data (halflife days), but execute based on intraday data (such as one min bars)?
Thanks for help.
Hi Anon,
ReplyDeleteIf you do not hold overnight positions, then you can compute halflife just based on intraday prices. Otherwise, it should be based on daily closes.
Ernie
Hi Ernie,
ReplyDeleteIs IB historical intraday data already dividends and splits adjusted? (such as one minute bars)
Moreover, Is intraday data from http://www.tickdata.com/ already dividends and splits adjusted?
Or we need to use information they provide, and do it ourselves.
Thanks very much.
Hi Anon,
ReplyDeleteIB historical data is not split/dividend adjusted.
tickdata.com will provide you with a splits/dividend file and an adjustment program so that you can apply these adjustments.
Ernie
Hi Ernie,
ReplyDeleteMay I ask if Yahoo! Finance End of Day adj close price reliable?
Do you compare it with tickdata.com?
Could we use yahoo data for Cointegration test?
Thanks very much.
Hi,
ReplyDeleteYahoo uses csidata.com data, and that is split/dividend adjusted, so I would trust them.
I use csidata for most of my backtest.
Ernie
Hi Ernie,
ReplyDeleteSo you use intraday data from tickdata.com and End of Day data from csidata.com?
Thanks for your time.
Hi Anon,
ReplyDeleteThat is correct.
Ernie
I personally found that Jeff Augen's books on options trading is more practical and mathematical. Hence, can be applied into small scale algo trading.
ReplyDeleteThanks for the tip, Swiss Dragon.
ReplyDeleteErnie
I went to tickdata.com to check their prices, and I have to say that the data I've been buying from Pi Trading is a lot cheaper. I'm not sure if their data is 100% as good as tickdata or some of the other data providers, so maybe others can chime in with some observations.
ReplyDeleteIt just seems that if you're paying over $5-20k for historical data when you could be getting it for free from IB or for <$200, this $5k data better be some Biff Tannen style data from the future, and it better come from someone who looks like an old version of me so I know that it's real.
- Brendon
Brendon,
ReplyDeleteSome other reader has suggested kibot.com for tick data as well. It looks very professional, and split/dividend adjusted.
Ernie
hi Ernest,
ReplyDeletelong time reader and a big fan of algo trading. For someone just starting out could you provide some idea how scaleable a one-man quant shop is? Without giving away specifics how many distinct strategies do you run? And how much could I realistically manage on my own? I ask this as i have some promising ideas but i don't want the hassle of hiring others. I mean, would I have to cap out at 1mill, 30mill etc..? Anyway, keep up the excellent posts!
Hi Andrew,
ReplyDeleteI believe it is quite possible to run up to 10 strategies by one person, though most people would be very happy with 3-5 consistently profitable ones!
If a fund is earning, say, $1M /year in fees, I can't see why the sole GP should not hire or partner with someone so as to provide business continuity in case of personal emergencies.
Ernie
This comment has been removed by the author.
ReplyDeleteHi Ernie,
ReplyDeleteRegarding cointegration for pair trading, how many days of data do I need to run cadf test ?
I am trading pairs with holding period of up to 20 days
For example, I ran cadf test for STI/CMA pairs and the results are:
with 1 year daily data, more than 95% probability that the pair is cointegrated.
with 3 months of daily data, less than 90% (not cointegrated?)
with 1 month data, the pair has more than 95% probability that it is cointegrated.
So, which data length should I decide? Thanks so much
Hi sg,
ReplyDeleteFor cointegration test, one year of daily data is minimum.
Ernie
Hi Ernie,
ReplyDeleteand when testing cointegration with intra day data? which data lenght should be used?
thanks
@Martin:
ReplyDeleteYou posted an interesting question..
I am interested to know the answer too.
Are you doing intraday pair trading, Martin? or, some HFT stuff?
@Ernie:
Thanks for the answer.
I have another question, if you don't mind, since I just bought your book...
1. The plot(z). Is Z basically the spread between stock #1 and stock #2?
2. I read somewhere about Engle-Granger and Johansen test. What are these 2 tests for (and compared to ADF) ?
Thanks!
@Martin,
ReplyDeleteIf you want to know if intraday data is cointegration, you have to test it on each day separately, or use aspecial method to concatenate all the different days data together with no gap (just like futures continuous contracts backadjustment).
@sg,
plot(z) on which page? Using Z means Zscore, or the residue from a linear regression fit.
Engle-Granger and Johansen are both cointegration tests for multiple time series. ADF is for testing stationarity of a single time series only.
E.
@sg:
ReplyDeleteI'm doing a research paper on this subject in the Brazilian market.
@Ernie:
thanks for the asnwer.
im using 5min data without adjustments. can you explain a little what would be the problem of using this kind of data for pairs trading? and why is it different from daily data?
Martin,
ReplyDeleteIf you are looking for cointegration on a daily basis, there is no need to use 5-min data. It won't help.
But if you are trading strictly intraday (exiting at close), then the overnight gap may cause the cointegration overnight to disappear which is irrelevant to your model. You need to adjust it away.
Ernie
Hi Ernie,
ReplyDeleteI just want to ask whether you recommend using market orders to hit both legs of a pair trade or whether that is too risky. I'm afraid the liquidity would suddenly drop away. But then again, even if I quote a limit price on one side and use the market side to hedge, there is also the same risk, albite on one leg only. do you think its recommended to use market orders on both legs and just hit the market price?
Ernie, from what i understand the problem with intraday data is if I exit at market close.
ReplyDeletebut if i use intra day data just for open and close positions and still carrying them overnight if the divergence persists(like as with daily data), then there is no need to adjust the 5 minute data,am i right?
btw, im not considering this HFT.
thanks for your help! And your book has a very good material, learned Matlab practicing with it.
Hi Anon,
ReplyDeleteI recommend you use limit order on one leg and market order on the other.
Ernie
Martin,
ReplyDeleteIf you do carry positions overnight, there is no reason to use intraday data for cointegration tests.
You can use intraday data merely to backtest entries and exits.
Ernie
@Ernie,
ReplyDeleteThanks for the answer
You asked me which page is the plot(z) on. It is on page 130.
Is it like the spread of Stock A and B?
Also, another question...
For HFT type trading, what kind of time frame the data we need?
There should be some kinds of cointegration that we can take advantage for HFT data?
@sg,
ReplyDeleteYes, the z on page 130 refers to the spread GLD-1.6766*GDX.
For HFT, you need to purchase tick data (trades and quotes). Cointegration is irrelevant at that timeframe.
Ernie
Great blog !
ReplyDeleteThis comment has been removed by the author.
ReplyDeleteErnie, can you explain me the similarities and the differences in the results between testing for cointegration in intraday and daily data, specifically for the pairs trading approach that we talked earlier.
ReplyDeleteI'm a novice in econometrics.
thanks
Martin,
ReplyDeleteIf you don't hold positions over night, then you want to be sure that mean-reversion happen during the trading hours. In this case, you should test for cointegration using intraday data. But you can't concatenate such data from different days without back-adjustment due to the overnight gap. (See my previous comments.)
On the other hand, if you do hold positions overnight, than you just need to use daily data to test for cointegration, since all you care is that mean reversion happens, no matter how long it takes. Testing with intraday data does not buy you any additional statistical significance.
Ernie
one last question. when we do hold positions overnight. I understood that intraday data do not bring additional statistical significance, but using intraday instead of daily data invalidates the cointegration test for that pairs trading purpose?
ReplyDeleteThanks Ernie.
Martin,
ReplyDeleteNo, using intraday data to test for cointegration doesn't invalidate the test if you intend to hold overnight.
Ernie
Hi Ernie,
ReplyDeleteThanks for the great blog.
Not sure if you already discussed but I would like to know your opinion on using stop orders on IB, do they get executed efficiently? what is a typical slippage in pips in FX in these cases?
Btw, do you have expected date for your second book? very much looking forward to it.
Thanks again!
Hi Quijano,
ReplyDeleteThanks for your kind comment. My new book is due probably in 2013 Q2.
Slippage in IB stop orders depends of course on the exact pair. For EURUSD, it is small, 1 or 2 pips. For EM currencies, can be more than the BA spread.
Ernie