Tag Archives: Loss

probability of a loss secret sauce tricks of Renaissance Technologies HFT masters


Sweet Mama. This is the closest I have gotten to this secret sauce eof Ren Tech black swan risk and probability of a loss tricks

From someone who knew all the founders of Ren Tech:

I have known Jim Simons, Bob Mercer and Peter Brown since 1965, 1974, and 1979, respectively.  Renaissance has also hired senior researchers who had formerly worked for me for years.  None of these people has ever told me anything about Renaissance’s investment strategies.  My observations below have been obtained entirely from publicly available records.

In particular, the core strategy is publicly known.  It’s the details that are proprietary.  There are millions of details, and they are essential to the performance.  However, the question was about strategy, so that is what I will try to answer.

The core strategy is portfolio-level statistical arbitrage carried to the limit and executed extremely well.  Basically, portfolios of long and short positions are created that hedge out market risk, sector risk and any other kind of risk that Renaissance can statistically predict.  The extreme degree of hedging reduces that net rate of return but the volatility of the portfolio is reduced by an even greater factor.  The standard deviation of the value of the portfolio at a future date is much lower than its expected value.  Therefore, with a large number of trades the law of large numbers assures that the probability of a loss is very small.  In such a situation, leverage multiplies both the expected return and the volatility by the same multiple, so even with a high leverage the probability of a loss remains very small.

The general properties of the strategy can be deduced from the statement of Renaissance for the Hearing of the Senate Permanent Subcommittee on Investigations, dated July 22, 2014.  [https://www.google.com/url?sa=t&…

Renaissance collects “all publicly available data [they] can that [they] believe might bear on the movement of prices of tradable instruments–news stories, analysts’ reports, energy reports, crop reports, weather reports, regulatory findings, accounting data, and, of course, quotes and trades from markets around the world.”

Their models “use this data to make predictions about future price changes.”

The hearing was specifically about the Medallion fund, about which the statement says “The model developed by Renaissance for Medallion makes predictions that are profitable only slightly more often than not.”

With these properties, there were two reasons that Renaissance would like to have a call option on the portfolio that it has designed: leverage and protection against Black Swan events.

Leverage is needed because, unleveraged, the rate of return of the portfolio is low.  However, because the volatility is much less than the expected return there is no limit to how high the leverage could be without increasing the probability of a loss, at least according to the models.  Through years of use and refinement, Renaissance knows that its models are very reliable.  However, they also know that there is always the risk of something happening that is not covered by the models, in particular something that is outside prior experience, which is called a “Black Swan” event.

Thus, a call option is ideal: it can provide high leverage and can provide protection both against the very low probability of a loss greater than the option premium and also against the unknown probability of a possibly catastrophic loss due to a Black Swan event.

We know all this because these are the business reasons for Renaissance accepting Deutsche Bank’s proposal of barrier options.  Basically, Deutsche Bank, and later Barclays,  sold the equivalent of a call option to Renaissance on the reference portfolio that Renaissance designed.

Of course, writing an uncovered call on the Renaissance portfolio would be equivalent to betting against Renaissance at high leverage, which would seem to be a foolish thing to do.  The banks covered these options by buying all of the securities in the portfolio.  Thus the bank’s position was equivalent to a covered call.  In other words, the banks’ profits and risks were essentially equivalent to writing a put option, which is a bullish position.  Because the volatility was very low the probability of a loss for the bank was low and the probability of a loss greater than the option premium was even lower.

Except for the Black Swan risk.  The probability of a Black Swan risk is unknown.  Part of the premium paid by Renaissance and earned by the banks was equivalent to insurance against Black Swan risk.  I don’t know if the amounts of the premiums were publicly disclosed.

There were many more details in the statements and the testimony at the hearings.  However, discussion of further details would detract from the important points that I have made above.  In particular, the hearings themselves were about tax issues not about investment strategies.  Renaissance explicitly asserted, under oath, that its “models do not factor in tax rates when making trading decisions.”  Therefore, tax issues, although they might be very important, are not part of the “investment strategy” at least as reflected in the models, so they are outside the scope of this particular discussion.

[Edit (added in answer to a comment):  The reference portfolio was highly dynamic.  There were thousands of  trades per day.  To accomplish this, the banks gave RenTech’s computers  direct access to execute trades through the banks’ trading desks.

This  arrangement was part of what created controversy about what should  be the proper tax treatment for this particular case. However, I am not a  tax lawyer and will not try to analyze those issues.  However, if you  want to hear more details on the automatic execution of the trades, and  questions about how much human interaction was present, that is all  discussed in the live testimony before the subcommittee: [Hearings| Homeland Security & Governmental Affairs]

I have copied this in case the Quora link disappears which is from


Notes from Senate hearings include:



Copy attached just in case that disappears

STMT – Renaissance (July 22 2014)2

Latest videos from the legend Jim Simmons

NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

Demo of live crypto currency positions profit and loss algo trading


I just did a 38 minute on digging into how certain positions can be profitable or losing money, I go so deep here I need to track a way to close out the positions for profit and loss measurement. The logs I generated are very good to see the life cycle of the position. You see see if position goes profitable and see how fast it can go negative. This is the advantage of using these type of logs. In fact, I may do a day webinar for my Elite members for this.

Here is the video

In the Trenches with Crypto Currencies video replay


NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

Minimize loss with linear regression for algo forex trading

True range forex trading market random performance video

I refer to linear regression below and in the long video below  There are 2 videos here on 3 trading sessions from this algo trading system I am working on. First is the long video of 1hr 50 minutes which explains each order text file my system generates. Do remember I am using Dukascopy JForex 3 which is the current version. I can also report that my system had recent problems where it hung after 5:15 earlier this morning. My local time is Eastern Daylight Time which is same as New York.

When you watch the videos, the key is to minimize the number of losing trades when you can take on new trades. It is imperative to measure the trend of bars before taking on the trade. I am using linear regression with my other 3 indicators in parallel for entry. Now there is a 4th but it seems checking for linear regression on each bar on 40 plus subscribed instruments will make your system hang like mine did. Just also remember I am using 14 gb system on Windows 10. I know but don’t ask. I plan to fix that with a new Linux based (eg. Ubuntu)  server with hopefully 64 gb system. I am looking at Dell t3610 tower which has good rating.

Anyhow, it is imperative to reduce or even eliminate those losing trades. I figure the linear regression (or same as trend line for technical analysis traders) helps here. Also, for the Jan 5 testing I was measuring 20 bars (each is 1 minute) but it maybe should be reduced to 5. I have certain flatline conditions which can mess this up with 20 bars. This is all theoretical of course but will try this Monday on the next trading day. I also highlight ways to improve the efficiency on the programming side in the 1hr 50 minute video below.

When you watch the video on the losing strategies, it needs to be understood the Dukascopy report does not distinguish between long and short trades. I also messed up in the videos of thinking short going negative would be profitable. This does not appear to be the case in the final demo reports. Who knows? We need to find ways to improve our trading profit potential.

More to come but I will expect to reduce the number of bad trades with improved linear regression checking for 5 bars instead of 20.

I do highlight in the long video that typical human greed can kick in as the positions/order life cycle can hit peaks before it closes/exits. It is very tough to build a systematic approach to the many trading conditions which are very random. Among my groups are the chatter about technical indicators like Bollinger Bands will help. I have found instances watching 1 minute price lines move way to fast before the Bollinger Bands catch it. I found once instance a currency pair lost 0.50 USD in just 6 bars. This could be virtually in seconds so these type of indicators may not be useful as one would think. I think it is quite hard for a human to track not just 40+ forex pair instruments at once , but it is very difficult to track every minute data bar. Just remember, my order text log file does track this as demonstrated in my long video.

There are way too many factors to consider when finding exit conditions because everything is so random or it just moves way to fast. It seems to best approach is to use the positions profit & loss (P&L) in US dollars is the best indicator out there. The hard part is to set targets on when to close but use stats to set that. I may have enough data but you want to ensure that you can let the profit potential run. As some examples clearly show they can do real well as in $7 on minimal amount traded is quite spectacular if you ask me.




Most stupid forex exit indicator could work? Average true range?


NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

Cross-Asset Quants Are Facing Worst Loss in a Decade

Cross-Asset Quants Are Facing Worst Loss in a Decade

Just like hedge funds, there are too many udnerperforming participants who call themselves ‘quants’. Sort of like your Uncle Joe and his frikkin dog.


NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

#statoil suffers first adjusted #loss

suffers first adjusted

[igp-video src=”https://quantlabs.net/blog/wp-content/uploads/2016/07/statoil-suffers-first-adjusted-loss.mp4″ poster=”https://quantlabs.net/blog/wp-content/uploads/2016/07/statoil-suffers-first-adjusted-loss.jpg” size=”large”]
#statoil suffers first adjusted #loss

NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

Linux CPP Demo of 2 futures contracts with basis and net gain or loss

Linux CPP Demo of 2 futures contracts with basis and net gain or loss

This our first crack at this. Comment at Facebook or Youtube to let me know if you see errors. Thanks

Get source here via my Quant Elite membership

Join my FREE newsletter to learn more about future with algo trading systems

NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

How to use statistics and probability with normal distribution for price movement analysis and exposure to reduce risk of trading loss?

How to use statistics and probability with normal distribution returns for price movement analysis and exposure to reduce risk of trading loss?

There is a video below!

Can you predict which days will be up or down for trading potential? If not, see below using stats and probability as those days are not much as it appears. There are many things against you. What is below is sort of mumbo jumbo but this is a crust of my upcoming analytics signalling service. This could be scary if you fully don’t understand so see the last two paragraphs.

Sometimes, your bell curve may show fat tails which could spell out extreme profit potential or high risk potential:

Fat Tail, Risk Budgeting, Factor Analysis & Stress Testing


In Excel, you could use Data Analysis add in with Descriptive Statistics for summary stats. Use returns for input.

Mean is average return of index (i.e. S&P 500) over historical period of asset your analyze. It should be always be positive if you analyze S&P since 1960s.

Median is middle of data set. If median is higher than mean that says the average is more positive over the mean or opposite would be more extreme if negative.

Standard deviation (SD) is dispersion of the distribution data. i.e. 68% of standard deviation 1 lies within the mean. Could also calculate normal vs actual with percentage for each. 68% of trading days that lie within SD 1the mean of the trading distribution. 95% will be for SD 2 and 99% of SD 3.

Kurtosis and skewness means if data is normally distributed. Kurtosis is a measure of how the data has peaked. High skewness can show the data could peak more than usual within a normal distribution with fatter tails. If you have a negative skewness, that means the fat tail are negative which means than the positive tail. Extreme negative movements are more likely to happen than positive.

Probabilities for new bins of frequencies within the distribution table. Thus the frequency of ranges from within the distribution table

=# of trading days in interval/total number of trading days in distribution table (in formula cell, press F4 for $ around on denominator to fix cell). This is represented in a cumulative representation.

You could also filter returns to understand subsets of data. We could look at an average of a subset.


Average return is calculated is the same as mean.

Want to understand the negative and positive of the returns of each average return with Data->Filter

If you calculate the frequency of positive vs negative, you could estimate market direction with probability. Predict average return when S&P 500 will go up when you long for x days, you forecast the daily return of this period. ****You will soon see that the percentage the market goes up is slim with negative days, transactions costs, and volatility against you. The forecasted’ “Average return” tells you this.

Also calculate the standard deviation (SD) returns as well with SD of 1-3 above the mean. Calculate the upper and lower of each SD above/below the mean. You could find the average number of days that lie within these ranges.

In summary, be aware:

In essence, since S&P 500 moves less at the extreme, it is harder to generate revenue with other assets. In OTW, our trading opportunities are minimal when you factor in AVERAGE returns on a daily basis when volatility is low.

*** You can also say that there are greater risks in the extremity of the tails of the distribution. They can be more probable than you think so there could be days where you could get wiped out in your account. There could be for example of 1 day per 100 where the market moves more than 3% which could wipe you out if you have the position the wrong way.

You could apply this logic over any period on any asset.

Join my FREE newsletter to find out when this new signalling process gets built to protect your trading capital



NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!

UBS Rushes To Make Changes After UBS Trader’s $2.3 Billion Loss compliancesearch.com

UBS Rushes To Make Changes After UBS Trader’s $2.3 Billion Loss compliancesearch.com

By Courtney Comstock The fallout from the UBS trader who lost $2.3 billion of the firm’s money betting on index futures is grim. Turns out, he…

One does NOT lose $2.3 Billion in a single day’s trading. It takes months to accomplish.
And this doesn’t go unnoticed by management. Traders do have $$$ limits, and lets not forget end of day reports.

The trader and his line of management were most likely aware of this disaster, and attempted to recoup the loses, but the dice were against them.

They all should be prosecuted

Some one rolled a 7, and they lost.

UBS’s Sr. management must ensure that the appropriate checks & balances are in placed and followed.

Happy Lovely


I think you’re correct in saying others had to be involved. Even if the trader was booking paper only offsetting positions with extended settlement dates (my understanding of the scheme) those should still cause a “break” in the back office. And if he was able to pull this off on his own management should be fired for stupidity.



NOTE I now post my TRADING ALERTS into my personal FACEBOOK ACCOUNT and TWITTER. Don't worry as I don't post stupid cat videos or what I eat!