Using Information Theory To Determine Bar Type & Frequency (Part 1)
My favorite bar type is based on information, not time.
Important: There is no guarantee that ATS strategies will have the same performance in the future. I use backtests and forward tests to compare historical strategy performance. Backtests are based on historical data, not real-time data so the results shared are hypothetical, not real. Forward tests are based on live data, however, they use a simulated account. Any success I have with live trading is untypical. Trading futures is extremely risky. You should only use risk capital to fund live futures accounts and if you do trade live, be prepared to lose your entire account. There are no guarantees that any performance you see here will continue in the future. I recommend using ATS strategies in simulated trading until you/we find the holy grail of trade strategy. This is strictly for learning purposes.
“Information is the resolution of uncertainty.”
-Paraphrased from Claude Shannon’s seminal 1948 work “A Mathematical Theory of Communication”.
It’s 9:29am. Your plan is set: fade the overnight range if liquidity stays thin.
One minute later, at 9:30 and a heartbeat, boom—1,200 contracts hit the bid, price free-falls 25 points, social media lights up with a surprise CPI beat, and you need a new plan.
What happened?
In the span of a single breath, the market goes from monastery silence to Kid Rock concert. Every tick suddenly matters; every second you hesitate feels like a mile of slippage.
Fifteen minutes later the frenzy subsides. Volume trickles back to its pre-open rhythms. You lean back, glance at the clock—9:45—and realize something obvious:
The market’s heartbeat isn’t measured in minutes; it’s measured in information.
This observation is exactly what information theory formalizes. It’s the science of surprise versus noise—of figuring out how much new insight each trade, headline, or order-book entry really carries, and how to process it in a meaningful way. For price charts, the most meaningful way boils down to bar type and the frequency of bar formation.
What I’m going to share with you today is a way to keep the surprise, dump the noise, and match how markets actually move with the use of a specific bar type. Scroll down for downloads (C#).
Using Information Theory To Maximize Surprise & Reduce Noise
Think of every data point—trade, tweet, note, earnings headline—as information.
Information theory asks two simple questions:
How surprising is that data point?
If you already expected it, it carries little information. If it shocks you, it carries a lot.How efficiently can we move that surprise from A to B without it getting garbled?
The more noise (latency, bad quotes, rumors) in the line, the harder this is.
From those two questions flow the big ideas for trading:
Surprise is valuable. A coin that almost always lands heads tells you almost nothing when it lands heads, but a lot when it lands tails. Big moves that break the “expected” range matter more than the everyday wiggles.
You can’t compress past the surprise. If market data are already predictable, further “edge squeezing” is impossible. Over-optimized strategies often fail because they’re trying to extract edge that isn’t there.
Every channel has a speed limit. Push data faster than the line can handle and errors explode. There’s a point where adding faster time-frames just injects noise and slippage.
Why does any of this matter?
Cleaner statistics. Equal-info bars tame volatility clustering, so back-tests stop “hallucinating” edge.
Faster signal latency. Indicators flip sooner during regime shifts because bars form rapidly when surprise is high.
Noise avoidance. In lunch-hour churn the bar clock slows, automatically throttling over-trading.
So from an information theory perspective, we’re looking for a way to interpret the market based on bar type and the frequency of bar formation. The former tells you how to calculate each information packet, the latter tells you how fast or slow those packets are formed.
Time Vs Information
Market information doesn't flow at a constant rate based on clock time. Sometimes it rushes in like a flood (high volatility periods), and sometimes it's a mere trickle (low liquidity periods). Traditional time-based charts completely ignore this reality.
Most traders default to time-based charts—1-minute, 5-minute, hourly—without questioning if these arbitrary divisions match how markets actually move. This is also the most common measure used for historical data—OLHC per minute—so it makes sense from a simulation perspective as well.
When I first started out, I actually used 36-range bars. I found this particular bar frequency to produce better results than any other in the backtest and set about creating numerous strategies without forward testing them. This newsletter documents the entire discovery process. Two years in, after being impaled by one too many backtested unicorns, I settled on minutes because the backtest produces more reliable results in minutes, not because it provides trading strategies with the best information.
But, I've always felt there was a more optimal bar frequency out there—one that isn’t hindered by the limitations of the simulation and the historical information we feed it. One that truly captures the market's natural rhythm. A bar type that matches how the market naturally processes information.