Turning to data for a trading edge · Dave Bergstrom, quant trader

EP 103: Escaping randomness, and turning to data for an edge w/ Dave Bergstrom

On this episode, I’m joined by a quant trader who works at a high frequency trading firm—though you might be surprised to hear, he started out on the same path that many retail traders do—his name is; Dave Bergstrom.

The thing that makes Dave unique from most traders who’ve been on this podcast previously, is how he uses data-mining techniques to develop trading strategies. Though data-mining, in trading, often has a negative connotation attached to it, Dave believes this stems from bad practices and poor evaluation of methods.

In addition to the above and ways to reduce curve-fitting, we talk about escaping randomness, learning to write code, Dave’s three laws for strategy development, setting expectations and plenty more.

--

[schrodingerscat]

this guy has walked on the exact same path I have. I really wish I had this video years ago when I was first introduced to trading. This is the way. Coding was on my bucket list for years, then I learned how to backtest and realized how much time was wasted manually analyzing data

[zzz_ttt_]

for data mining the best choice is python ... you can learn it fast. I use C++ since 1997, but when it comes to fast prototyping python beats C++ by light years. I learned it just to be able to datamine. In 100 lines of code I could do analisys, graphs, everything. What you need is a database .. which you can build for yourself by logging data from a platform. I did a lot of cheap solutions. Having access to a good data source is the key. I used Tensorflow to make some neural network to react for pivot points, but it worked pretty poorly :)). It is pretty big amount of work after all.

[TehBr]

3rd time watching this episode and still getting value out of it - cheers.

[riquenosis]

Wow great but difficult interview for both parties. Bro you definitely extracted water from a centerblock on this one... he wasn't giving up much at all... yet did give a few generalities.

[elguille]

Great interview Aaron. Keep up the amazing work, very much appreciated!!!

[tipswontjumpship]

got my golden nugget! good looking out Aaron! your beast bro!

[jamesjohnson]

His website is buildalpha.com it helps you build trading systems without knowing how to coded. For example it offers 5000 entries and 5000 exits . Once you come up with a system you like just print the code for tradestation, python, multicharts etc.... This what the interview was missing. Deeeep

[gandaslifeadventures]

Anybody have any strategy development books for recommendations? For any programming language. I currently code with python. Using mostly anaconda.

[kotsios]

Please make more and more and more i love them

Sooo much educational VALUE here :)

[bobbybax]

I'm a mean reversion trader and it requires more discipline in that its critical to realize when you are wrong.

[modelworkzseo]

variance testing; that's the gold in this interview: 37:00

[jetal]

what does he mean saying using open high low data for tech indicator. Is he talking about pre market high lows or first 15 min high or lows?

[anothernumber]

lol do any of you guys remember the scene on the sopranos where they were trying to inflate the value of webestics and then what happened after? Lmao

[honglee]

wish the janeky bino options scams would stop advertising

[josouk]

So many generalities ! That is the problem with so many of these interviews. Really doesn't want to give anything away ! For once can we get down to specifics !!!

[DailyStockPicks]

Every word is genuine. My trading is based on his philosophy. I trade 100% transparently on my channel. You can see it in front of your eyes, everyday in live market, the practical aspect of what he covered in his interview.

[baibhavtiwari]

Even you have done your all 1000 trades in bear market and you are always short and you are right 80% of time. Does it mean you have an edge ? You will be wiped out when bull market will come. You should define your strategy considering all type of market condition.

[pahriz]

This guy was speaking in such generalities, I couldn't listen anymore.

[HitAndMissLab]

He's not a really full size quant. He's sort of self-thought, although he thought himself about 5% of what graduate quant knows. So his skill-set is far bellow average.
Discretionary traders and quants can't be compared because human brain doesn't work like computer. Human would always see full context of a chart, while computer can follow only 2-3 parameters. That's why human needs less then 50-100 samples and computer needs many thousands. And, that is why human can trade several different patterns together and across multiple markets, while quant would need 5 separate systems. Practically quant has to wait 1, 000 trades to scale up with one more contract, which might take a year, while some of discretionary traders go from $500 to $17, 000 in two weeks.

[jstfunx]

Total waste of time. And BTW, you guys don't know what data mining is. You should be pretty embarrassed. The term is the old buzz word for what they have re-branded as Data Science. Data-mining uses statistical algorithms such as clustering, neural networks, and decisions trees to find patterns and relationships in data. It is different than just a trading algorithm which may simply encapsulate your trading rules and not use any statistical methods.