Build Alpha Trading Software creates, tests and codes trading strategies with the click of a button. Demos: https://xitter.cf/t.co/kF5Niev0Wg
Boca Raton, FL
Joined January 2017
- Tweets3.3K
- Following0
- Followers8.4K
- Likes866
Build Alpha retweeted
This man is gathering all the infinity stones and been COOKING!
Definitely tuned in for this. Thank you for sharing!
Build Alpha retweeted
I really like the football analogy, it perfectly explains how conditioning on different events can uncover useful variations and build genuinely interesting systems off them. Please drop more sauce like this 👏
Would you consider adding more events/features to the signal search?
Build Alpha retweeted
Edge leakage since 2017 and you all still dont understand. This is so good.
Markets have memory. Most trading signals don’t.
The same signal can mean something completely different depending on what happened earlier in the session.
And what happened earlier changes what happens next.
Think of a football game's score. 21–17 tells you the state of the game. It matters if the team down 17–0 has just scored three straight touchdowns, or was ahead 21–0 and is collapsing.
That’s the idea behind event-conditioned signals. A trading signal conditioned on an earlier event.
Examples:
• Yesterday’s last hour was strong → does today’s first hour follow through?
• Opening range closed higher → does the last hour tend to rally?
• Price was above VWAP at noon → does that strength persist into the close?
• Today’s opening range is unusually large → are breakouts more likely to continue?
I've added a ValueWhen function that stores what happened at a specific time or event, then we can use that information later or even the next session.
Available in the custom editor and the prompt to signal feature.
This greatly expands the types of signals and strategies possible.
I put together 168 ready-made event-conditioned signals covering:
• Last-hour behavior
• Opening-range behavior
• VWAP checkpoints
• Open / ORH / ORL relationships
• Body-to-range ratios
• High/low-of-day structure
Then you can mix them with Build Alpha’s other built-in signals and/or your own custom signals and let Build Alpha + AI test which combinations actually survive validation.
The more context your strategy has the more robust it can be.
Build Alpha retweeted
I’m still learning this amazing software Build Alpha. This morning I discovered Chart with AI, I think this feature deserves a YouTube video
Build Alpha retweeted
I badly need to set this up
The best trading research workflow is one you don’t have to sit and supervise.
That’s the idea behind Sequence Workflows in Build Alpha.
Instead of:
Run a broad search on 10 years of data
Wait, then manually
Take the survivors and test the last 3 years
Wait, then manually run
Run Robustness Tests like Monte Carlo
Wait, then manually
Test on another date range as a final check
…you can now configure the entire research pipeline once and let Build Alpha run it end to end.
Each round can have its own:
• Date range
• Performance filters
• Robustness filters
• Advancement criteria
Example:
Round 1 → mine until you find 2500 candidates
Round 2 → first out of sample test
Round 3 → Monte Carlo Permutation
Round 4 → Noise Test, Vs Random
Round 5 → final out of sample test
You can also tell it:
“Advance when 200 strategies pass, OR after 3 hours, whichever comes first.”
No more manual babysitting. No more coding.
Set the sequence. Hit start. Come back and see either
· surviving strategies
-or-
· a clear answer where things broke down
Both are actually useful and save a lot of time.
Build Alpha retweeted
That’s powerful, leveraging the system to do the deep digging without the need to supervise. Thank you!
Build Alpha retweeted
How can you tell if your exits are overfit?
System traders can overfit both entries and exits. Hard to tell which but we need to check both.
Monte Carlo only changes the order of existing trades. It preserves the profits and losses produced by the exact exit settings from the backtest.
Build Alpha’s Randomized Monte Carlo Test challenges the settings themselves:
• Keep the entries fixed
• Randomly vary exit parameters on each trade
• Repeat 1,000+ times to get a distribution
For example, a strategy using a 2 ATR stop and rolling 5-bar low as exits might change the ATR multiplier and rolling window for each trade.
No exit rules are added or removed. Just modified.
If most paths remain profitable, stronger confidence you didn't overfit the exits.
If profitability collapses, you're likely in for a rude awakening when you go live. At least you caught it first.
Your equity curve is one possibility. It shows what worked with one configuration. This test reveals how much depends on getting that configuration exactly right.
Stress test both your entries and your exits.
Build Alpha retweeted
If you're thinking about automating or still placing some trades manually to "hedge" your automated systems,
do this little experiment my mentor had me do.
Mark every manual trade in a spreadsheet for 3 months or a quarter. Do not cheat. Mark every single one.
1/
Build Alpha retweeted
Nice YT channel @buildalpha !
Build Alpha retweeted
This man single handedly changed how I see luck in trading
You can NOT eliminate luck from trading.
But you can increase your Luck Surface Area.
Luck Surface Area = the number of independent opportunities you give market noise to work in your favor.
A trader with:
• 1 strategy
• 1 market
• 1 timeframe
• 1 parameter set
needs a very specific combination to work. Unlikely.
A trader who diversifies across:
• Strategies
• Signals
• Markets
• Timeframes
• Parameters
• Entry/exit logic
• Alternative data
is giving market noise more opportunities to land in his/her favor. This also reduces dependence on any single outcome (generally a good idea).
This is one reason I built things like parameter diversification, alternative data, multi-market testing, portfolio simulation into Build Alpha.
Don’t try to predict every price fluctuation. Trust me, even with PhDs, machine learning and massive compute there is lower hanging fruit.
Build a process where randomness has more ways to help you than hurt you.
Build Alpha retweeted
Great article!
We all say the same thing about overfitting: the strategy was "fitted to the noise."
So test that directly. Change the noise and see if the strategy survives.
This is the entire idea behind the Noise Test.
Your price history has two parts. Signal, which is the part that repeats. Noise, which is the part that happened once and will never happen the same way again. Every strategy you build is fit to both, and your backtest cannot tell which one it learned.
So change the noise and re-run the strategy.
Here's how it works:
Pick how much of the data to adjust and by how much. Say 50% of the bars, by up to 25% of average true range. Opens, highs, lows and closes get nudged up or down inside that band. Same market, same trend, same volatility (and clustering) but a different roll of the dice.
Create 1,000 noise adjusted data series and re-trade the strategy 1,000 times.
Now you have a distribution instead of a single backtest.
How to read it:
PASS The noise adjusted results do not deviate from your backtest. Most (or all) stay profitable and the distribution sits meaningfully above zero. Your rules were tracking something real.
FAIL The noise adjusted results fall off a cliff. Your original backtest sits way above the distribution and the rest cluster near or below zero. No edge. You fit the one path history happened to take.
If adjusting the noise erases your edge, your edge was never there, and no amount of out-of-sample data would have told you, because your out-of-sample sample data has exactly the same problem: it is one path only.
Reminder: Passing is not proof of edge. It rules out one specific failure and there are plenty of others, which is why this test is only one of a dozen tests (all automate-able) in Build Alpha.
Build Alpha can also optimize your parameters across the noise-adjusted series instead of the single historical one, so you are selecting parameters that work across many versions of history rather than the exact history we lived.
Your price history is one sample. Your strategy should work on the others too!
Build Alpha retweeted
ICYMI here’s a summary of the quant topics covered this week.
Mostly centered around increasing risk adjusted returns by adding uncorrelated strategies (and the formulas/math behind it).
Give a like or follow. More planned for this coming week..
Four posts this week. One argument.
1. IR ≈ IC × √N
Edge comes from breadth, not strength. Fifty signals at IC 0.05 produce roughly three times the risk-adjusted edge of one signal at 0.10.
2. N_eff = N / (1 + (N−1)ρ)
Correlation caps breadth. As you add strategies, effective bets converge to 1/ρ and stop. At ρ = 0.30 you can never exceed 3.3 real bets no matter how many systems you own.
3. Sₛ > Sₚ × (√(2 + 2c) − 1)
What a candidate is worth is set by its correlation to your book, not by its own backtest. A 0.70 Sharpe strategy at 0.80 correlation takes a 1.00 Sharpe book down to 0.90 despite being a winner on its own.
4. Contracts = (Equity × Risk%) / (ATR × PointValue)
Fixed size is a leveraged bet on calm markets. ES ATR went from 20 to 200 in Q1 2020. Same two contracts, ten times the risk. Size based on volatility.
Here is what ties them together.
In all four, the number that decides the outcome is not on the strategy report.
Your Sharpe does not tell you your portfolio value. Your strategy count does not tell you your breadth. Your contract count does not tell you your risk.
Standalone metrics describe a strategy in isolation. You do not trade anything in isolation. Portfolio fit and proper sizing is more of the game than you realize.
Thanks for all the likes, retweets and follows this week.
