Bitcoin Edge Went 14-5 This Week. Here's Why It Worked, Why It Won't Always, and What We're Changing

I've been doing this long enough to know the most dangerous week a trader can have is a great one. You start believing 74% is who you are. It isn't. So here's the full report: what happened, why I think it happened, every hole I can find in it, and what we're changing so the alerts you get are the ones that have actually paid.

Every number below comes from our graded records. Nothing is cherry-picked, and the losing buckets are in here too. Live boards are on Bitcoin Edge and the 15-minute markets hub.

The scoreboard

Read it left to right and the story tells itself. Everything is green this week. Go out to 30 days and almost everything is red. All time, three of the four options-linked tools make money and oil gives a little back.

Two things to keep in mind:

• Gold, silver and oil fired 3, 1 and 4 times this week. Those cells are noise. A 3-1 week means nothing. Gold, silver and oil have all gone quiet since August, and they won only 13–16% of picks in August and September on tiny volume. I'll come back to that.
• Bitcoin Edge and the momentum rule are the only ones with enough volume this week to discuss. So that's where we're going.

Average net result per contract after Kalshi's taker fee, one contract per pick, graded against Kalshi settlement. The momentum rule is a pre-registered paper record, not graded picks.

The hot week, inside a rough month

This chart is the whole lesson. The running total bottomed at −$5.33 on October 14, the same week we're calling hot. The 14-5 record came from steady wins, but most of the money came on two days, October 15 (3-0) and October 16 (5-1).

And on October 14 the tool went 2-2 and still lost 46 cents. That's the first thing to understand about these markets:

Win rate isn't profit. A YES at 64¢ pays you 36¢ when it hits and costs you 64¢ when it misses. At this week's average price of about 64¢, you need to hit roughly 66% just to break even after the fee. We hit 74%. That 8-point cushion is the entire edge, and it's thinner than "74%" makes it sound.

Same idea in dollars: +8.5¢ on a contract that costs about 64¢ is roughly a 13% return on the money at risk for the week. If you go into next week expecting 13% a week, you're going to make bad decisions. The all-time version is +3.5¢ a contract, and that's the number I plan around.

Why it worked: four reasons

1. We plugged the leak on October 2

In late September, Bitcoin Edge kept finding "value" in cheap YES contracts, 15¢ to 30¢ shots that the model thought were underpriced. They went 3-31 and cost $5.17, which was more than the entire 30-day drawdown. Everything else that month was roughly flat.

On October 2 we stopped grading any pick that costs under 40¢. Those signals still get logged, they just don't count, and they don't alert. Since then, zero longshots, and the tool has been living where it has always made money: YES contracts between 40¢ and 70¢. All time, that's +11.0¢ per contract at 40–55¢ and +4.3¢ at 55–70¢.

I'll take credit for that one. The model didn't suddenly get smarter. We just stopped letting it swing at bad pitches.

2. Bitcoin sat in a box, and the box was well defended

All 19 picks this week were YES, mostly "bitcoin finishes the hour above X" with X close to the price. That's a contract that loves a calm, range-bound market. And that's what we got. Bitcoin opened Friday around $82,100 to $82,700, up only about 0.6% on the week (Yahoo Finance). It barely moved, even though September's CPI and PPI both came in hot.

3. The dealer-positioning backdrop helped

This is the part most retail traders never look at, and you should. Options dealers hedge what they sell. When a lot of open interest sits on strikes close to the current price, their hedging tends to lean against moves: they sell rallies and buy dips, and price gets pinned in a range. When price punches through a crowded strike, the same hedging can speed the move up.

Going into Friday's weekly expiry, Deribit's bitcoin puts were stacked at $78,000 and $80,000, with max pain (the price where the most options expire worthless) around $83,000 and bitcoin trading in between (Cryptoticker). That's a floor under the market and a magnet above it, the exact setup where an "above X this hour" read priced off options volatility tends to do its best work.

I'm being careful here. That's a likely explanation, not a proven one. We tag a dealer-gamma regime on our picks, but only 18 of more than 2,500 commodity picks actually carry the tag. That's one of the fixes below. Until it's wired up, this is a theory backed by public positioning data, not a measured edge.

4. The market agreed with us, fast

After the average pick this week, the price moved about 4.8 points toward our side. In trading terms, we beat the closing line. That's the best sign you'll ever get that a read is real and not luck. It also explains the first hole below.

Now let's poke holes in it

Here's the part most "look how great our picks were" posts leave out.

Hole #1: You couldn't have gotten those prices. Bitcoin Edge grades each pick at the market price when the signal fires. We tested what's actually available: across 2,100 strong signals in the last 14 days, the asking price about 10 seconds later was 4.2¢ worse on average, and worse 80% of the time. Some of that is our pipeline's lag. Most of it is the market moving our way, which is reason #4 working against us. Take 4¢ off and this week goes from +8.5¢ to roughly +4¢ per contract. Still good. Not 13%.

Hole #2: Nineteen picks is a coin-flip sample. The statistical range on 14-of-19 is roughly 51% to 88%. This week fits inside "a good run from a 58% tool." It doesn't prove we got better.

Hole #3: Every pick was the same trade. Nineteen YES contracts, same direction, same kind of market. That isn't nineteen independent results. It's closer to one view on bitcoin staying in its box, expressed nineteen times. When the box breaks, they'll lose together.

Hole #4: Our own robot couldn't fill. We run the 15-minute momentum rule live at small size ($3 max a position) on our own Kalshi account. In its first week it placed 23 orders and filled 2. Our order timing was fine: orders went out within two seconds of the :55 mark on fresh quotes. The problem was price. The bot's order price sat below the actual asking price, so its fill-or-cancel orders mostly cancelled. The ugly part: 16 of the 21 missed orders would have won. The two that filled split 1-1. That's a bug, it's ours, and the fix is below.

Hole #5: The momentum record is paper. 178 paper fills priced at the ask, with the fee included. It's honest paper, but until it clears the 500-window bar we set before the rule went live, it's a watch, not a product.

Hole #6: Gold, silver and oil are hiding behind a nice week. A combined 6-2 this week sits on top of a 13–16% hit rate in August and September and a pick count that fell from 100-plus a month to about 5. I wouldn't trade those alerts right now, and we're not going to push them.

The 15-minute lab: what's working, what isn't

Kalshi's 15-minute markets are where the platforms are pushing, so we track every window on bitcoin, gold, silver and WTI oil. Our 15-minute model is a research scorecard, not a track record. We publish it either way so you can see what we see. New to these contracts? Start with how Kalshi 15-minute markets settle and the fee math that decides whether a 15-minute strategy can make money.

The one rule that's paying: top-of-hour momentum

The rule is simple. Five minutes before the 15-minute window that closes on the hour, if bitcoin has moved since that window opened, take the side it's moving toward. Pay the ask and the fee. No model, no opinion. 126-52 since September 23, +4.05¢ per contract after fees, +$7.18 on one contract a signal.

Where it works and where it doesn't, on the full sample:

Look at that 80¢-and-up row. It hits 94% of the time and makes nothing. At 80¢-plus you're risking 80 to win 20 and paying a fee on top. That's the best example I have of why win rate is a vanity number. The money is in real moves during U.S. hours at prices under 80¢.

The buckets that bleed

Take our 15-minute model's side in every window and you lose about 1.5¢ to 3.5¢ a contract after fees. The details matter more than the average:

• Small disagreements don't pay the toll. When the model and Kalshi are 5–10 points apart, bitcoin, gold and silver all lose 2.5¢ to 4.4¢ a contract. The fee eats the whole edge.
• Our bitcoin 15-minute model has no edge. It loses in every disagreement size and every volatility regime. On bitcoin, the momentum rule is the play and the model is a reference.
• Silver longshots are a donation. Silver contracts priced under 20¢ hit 1 time in 38.
• The clock matters more than the model. Windows closing 6–10 a.m. ET (the pre-open scramble) and 7–9 p.m. ET were the worst of the day.

Where the 15-minute edge actually lives: quiet tape

This is the most useful chart in the report. When the model disagrees with Kalshi by 10 points or more and the market is calm, it makes money on gold (+10.4¢), silver (+5.5¢) and oil (+5.2¢). Same disagreement in a wild market and it loses on all three (oil −8.3¢).

The reason makes sense. The model prices off how much a market usually moves. In quiet tape, Kalshi traders overreact and the model is right. When a market is ripping on news, the model's sense of "normal" is stale and the crowd is right. Over the last seven days the 10-point-plus filter made money on all four markets, led by WTI (+8.5¢) and gold (+7.3¢), but over 30 days only WTI is positive. Calm tape is the condition. Without it, there's no edge.

The environment right now, and why next week could look nothing like this one

This is the part that should keep you humble going into the week.

• Bitcoin: about $82,000 to $83,000, chopping above the $80,000 put wall. The October 30 monthly expiry on Deribit has big call interest stacked at $90,000, $95,000 and $100,000, more than $4.5 billion between them, and Deribit max pain is near $78,000 (TokenPost). Expiry weeks can break boxes.
• The Fed: hot September CPI and PPI put an October hike back in play (Yahoo Finance). The FOMC meets October 27–28, with September PCE and Q3 GDP on October 29. Live odds are on our Fed rate tracker.
• Oil: WTI settled at $101.41 on Friday, up 10.4% on the week, and Brent touched $110 as attacks around the Strait of Hormuz cut flows (EnergyNow). That's wild tape, which is exactly where the oil 15-minute model loses. Don't overfit the calm-market results to this market.
• Gold and silver: gold around $4,166, down about 4% on the month and well off January's $5,608 record. Silver around $57, down about 12.6% on the month (Trading Economics: gold, silver). Metals have been in a downtrend with spikes, a different market from the steady one our metals edge tools made their money in last spring.

Put it together: the conditions that paid us this week (a pinned, quiet bitcoin and calm metals windows) are set up to be tested by an expiry, a Fed meeting and an oil shock in the next two weeks. That doesn't mean the tools stop working. It means we size down and let the market prove the box still holds.

What we're changing

We said the tools would run as self-improving loops. Some of that is running, and some of it was never finished. Here's the honest status and the fixes:

What's running: the 15-minute calibration jobs and the paper-rule grader run every hour, the weekly calibration refit runs every Monday, and commodity picks grade against Kalshi settlement on schedule. All of those succeeded every run this week.

What wasn't finished, and what we're shipping:

1. Published picks move to executable prices. Bitcoin Edge picks will be graded at the price you could actually get, ideally our live bot's real fill, not the signal-time price. Expect the record to look worse and be more honest.
2. The 40¢ floor stays, and the top end gets stricter. YES contracts at 70¢ and up hit 87% of the time but only make 1.3¢ a contract. Above 70¢ we'll require a bigger edge before a pick counts.
3. 15-minute alerts become conditional. An alert fires only on a 10-point-plus disagreement, in a calm or normal market for that commodity, outside 6–10 a.m. and 7–9 p.m. ET, and never on silver under 20¢. The bitcoin 15-minute model stops alerting. On bitcoin, the momentum rule does the talking.
4. The momentum rule gets a better filter. It favors setups where bitcoin has already moved 0.15% or more, prefers U.S. hours, and skips entries above 80¢. It stays in watch until it clears 500 windows.
5. The live bot gets fixed. It will price orders from the live order book at the moment of the decision, log every decision, and flag itself after three straight misses.
6. The data holes get closed. About 19% of 15-minute windows in the last week had gaps in our order-book recording, which is why some windows show up as "capture gaps" in the paper record. That gets fixed at the source.
7. The dealer-gamma tag gets wired to every pick, so next time I talk about positioning, it's a measured result instead of a theory.
8. Gold, silver and oil hourly and daily alerts go to watch until the model shows it can tell a steady market from a shock.

How I'd size around a hot week

This is education, not financial advice. It's how we think about our own risk.

• Never size up because of a hot week. Size off the long-run number (+3.5¢ a contract on Bitcoin Edge, less after realistic fills), not off 74%.
• Use one small fixed unit. A full unit goes to the buckets with a long, positive record: Bitcoin Edge YES at 40–70¢, momentum on real moves during U.S. hours.
• Cut to half a unit around catalysts: an FOMC meeting, a monthly options expiry, CPI or PCE day, or when a market is ripping on news.
• Zero in the losing buckets: longshots, small 15-minute disagreements, pre-open and evening windows, wild tape.
• If you use Kelly, use a quarter of it or less, fed with an honest edge. Our Kelly calculator does the math, and the Kalshi fee calculator tells you what the fee does to a thin edge.

Where to watch it this week

• Bitcoin Edge: hourly Kalshi bitcoin reads during options hours, graded against Kalshi settlement. New to it? Here's how to trade Bitcoin Edge in plain English.
• Bitcoin 15-minute: the momentum rule's live call for each top-of-hour window, plus the paper record.
• Gold 15-minute · Silver 15-minute · Oil 15-minute: the research scorecard, published either way.
• The 15-minute markets hub, crypto markets and commodity markets boards.
• The full graded history lives on our track record. How the engines work: our commodity edge methodology and Kalshi bitcoin vs. spot vs. options-implied.

I'll be at NEXTPredict in New York on October 22–23. If you trade these markets, come find me and tell me where I'm wrong. That's how the tools get better.

Trade responsibly.

How we measured this

• Window: seven days ending Friday, October 16, 2026 (Eastern time); 14 and 30 days on the same basis; all time since each tool's first graded pick (Bitcoin Edge since May 18, 2026).
• Graded picks: Bitcoin, Gold, Silver and Oil Edge picks are graded against Kalshi settlement. P/L is per contract, one contract per pick, after Kalshi's per-contract taker fee. Totals are labeled "net" and averages "per contract."
• Momentum rule: pre-registered on September 23, 2026, and recorded as paper fills at the asking price plus fee. It is not counted as graded picks.
• 15-minute research scorecard: every settled window from our model's first read, taking the model's side at the market midpoint with the fee applied. It is published either way and is not a track record. WTI's record starts September 28, 2026, when its price input changed.
• Fill slippage: the bot logs the live asking price about 10 seconds after each signal; the 4.2¢ figure covers 2,100 signals with an edge of 8 points or more over the 14 days ending October 16.

Sources: Yahoo Finance, bitcoin and ethereum prices, Oct. 16, 2026 · Cryptoticker, bitcoin options expiry Oct. 16 · TokenPost, Deribit calls cluster before Oct. 30 expiry · EnergyNow, oil weekly close Oct. 16 · Trading Economics, gold · Trading Economics, silver