An independent concept, built by a subscriber. Not affiliated with, endorsed by, or
published by Voltick.
Bot Ideas · SPY · 08/24
Pick an idea. It trades your levels — and shows you its receipts, wins and losses.
The idea
Fixed entry and exit logic, written down and versioned. You cannot edit it, which is the point.
Your settings
Size, daily stop, how many trades. Everything about you — nothing that changes the edge.
Receipts
Every idea ships with its record on your levels, including the ones that lost.
▲ Paper only11 sessions logged
Choose an idea
Reversion
Price drifts away from the Volt — the strike
with the most gamma pinned to it. This buys the move back toward it and gets out at the
next level standing in the way.
TriggerPrice stalls 0.30–0.75 short of the Volt and turns back toward it.
Not the first touch — the turn.
DirectionWhichever way the reaction goes. Approached from above and
held, it buys calls; from below and rejected, puts.
TargetThe next level in the path, exited $0.50 before it, because price
habitually turns short of the print.
ExitTrim at +20% and the stop moves to breakeven. Second trim if the
contract doubles. Otherwise a 5-minute break of structure ends it.
Stands downBefore 09:40, after 15:00, when the reward does not cover the
risk, or when the same level was traded in the last 20 minutes.
14
trades
36%
win rate
+$8.85
per trade
Rejection Fade
A level holds. Price pushes into it, fails to get through, and turns —
this trades the turn, away from the level.
TriggerPrice must reach the level's reaction band AND close back past the
rejection edge. Both, in that order.
DirectionAway from the level. Pushed down into it and failed means the
bounce is up.
TargetThe next level behind price, again at the near edge.
ExitSame ladder — +20% trims and arms breakeven, the double trims again,
structure ends it.
Stands downIf the level broke clean and has not yet failed a reclaim
twice, or if price already traded deep through it. A level that has been run over is
not a level that held.
12
trades
42%
win rate
−$1.40
per trade
Opening Range
The simplest of the three, and deliberately so — it ignores levels entirely
and trades the first hour's range break.
TriggerThe 09:30–09:40 range breaks, and price closes beyond it.
DirectionThe direction of the break. No opinion beyond that.
TargetNone. It rides until structure turns or the clock runs out.
ExitFlat by 10:30 whatever has happened.
Stands downEverything after the first hour.
11
trades
45%
win rate
−$3.52
per trade
None of these has proven an edge yet. 11–29 trades is far too small a sample to
separate skill from luck — at this size a coin flip can look like either. These numbers are
published so they can be checked, not because they are conclusions. Paper only until the
sample is real.
Yours to set
Capital per trade
Premium committed to one position
$600
Max trades per day
Hard cap, no exceptions
3
Stop for the day
Two losses and it stands down
2 losses
Exit style
Three measured profiles — not a slider
Notice what is not here: entry trigger, stop distance, strike selection, target.
Those are the idea. Exposing them as numbers lets you tune until the backtest looks
wonderful and the account does not — on a sample this size, that outcome is close to
guaranteed. Anything that changes the edge is a named, measured choice; anything about your
own risk is yours to set freely.
What it sees right now
Reversion · SPY11:15 ET
Flip767.00+3.99
Volt765.00+1.99
Surge763.00−0.01
Spot763.01
Standing down. Price is sitting on the Surge, which this idea does not trade —
it is live flow, not structure. The nearest level it will act on is the Volt at 765,
1.99 away, and the trigger needs 0.30–0.75. Nothing to do.
A bot that says why it did nothing is the difference between trusting it and
turning it off. Every session logs the setups considered and the reason each was declined.
Where these numbers come from. My own paper-trading harness, replayed over
11 SPY sessions (7–21 August 2026) against Voltick levels. One position at a time,
$600 of premium per position, a 20% stop, a trim at +20% and a second at +100%.
Option fills are modelled, not live. No real capital has traded any of this.
At 10–29 trades per idea, none of these figures is statistically meaningful yet — they
are published so they can be checked, not because they settle anything.
Why it passed · 11 sessions
123setups seen · 10 traded
It declined 92% of what it looked at. Every one is logged with
the reason, and scored afterwards on what it would have made.
crossedtradedno confirmat the bellother
Every reason, and what it cost
Crossed the level
96 · 78%
−$17
Price went straight through instead of stalling and turning. No setup ever formed.
Traded it
10 · 8%
—
Cleared every gate. These are the only ones that became positions.
No confirmation
7 · 6%
−$0
Reached the band but never turned. Waiting for the turn is the whole idea.
Open at the bell
5 · 4%
—
Inside the first ten minutes. It does not trade the open.
Cooling off
4 · 3%
−$115
Same level, too soon after the last trade there. One wobble is not two setups.
Reward too thin
1 · 1%
−$2
Target sat too close to pay for the risk.
Every decline bucket lost money. Scored after the fact, the setups it passed on would
have averaged a loss in all four cases — the cooling-off rule alone sidestepped
−$115 a time. Saying no is most of what it does, and so far saying no has been right.
Shadow figures are scored post-hoc on what actually happened after the decline. They cannot
feed a live decision and are not used to; they exist so a rule can be judged on evidence
instead of on argument.
And here it was wrong. Price turned up off 765 at 11:22 and ran nearly two points.
The grade said slippery — expect price to accelerate through — so the idea stood aside,
four separate times, at a level price respected all day.
This screen is what makes that visible. A signal you can only see when it fires cannot
be argued with; one that shows its refusals can be corrected.
What it's learning · 11 sessions
Every session adds evidence. When a pattern clears the bar, it becomes a new idea.
Nothing here is a recommendation. These are observations with their sample sizes attached,
ranked by how close each is to being answerable.
123 setups logged10 traded
Observations
CANDIDATE
The take-profit sat far too high
Peak contract appreciation reached +100% on 1 trade in 13 and +20% on 5. A
take-profit at the double is inert on 92% of trades.
Acted on. The trim moved back to +20% with the double kept as a second
stage. This is the only observation here that has changed a rule.
CANDIDATE
Slippery grades may be wrong when a level holds twice
On 08/21 the grip grade declined the same level four times — and price turned there
and ran nearly two points. Across all sessions the grade has been right far more often
than wrong, so this is not "ignore grip".
Proposed idea: trade a reaction off a slippery level once it has held
twice. Needs ~20 more slippery declines to test.
WATCHING
Saying no is where the money is
All four decline reasons scored negative after the fact. Cooling-off alone sidestepped an
average −$115 a time, and 78% of level touches never form a setup at all.
Suggests tightening entry further rather than loosening it.
Counter-evidence would be a decline bucket turning positive.
WATCHING
Account size is picking the strike
The decay cap is measured in dollars, so it binds harder as size grows. At $600 the
selector buys delta 0.45; at $3,000, on the identical setup, it buys 0.86.
Proposed idea: make the cap a fraction of capital deployed, so the
same setup picks the same contract at any size. No outcome data yet.
TOO EARLY
Which idea is best keeps changing
At 5 sessions, Reversion looked like −$28 a trade and Rejection Fade +$85. At 11 sessions
that inverted. Neither number was ever distinguishable from zero.
Per-trade spread is $69–$130 against single-digit means, so every 95% interval
spans zero. Roughly 60 more sessions before a $25/trade edge is
detectable.
Toward an answerable sample13 / ~62 trades
At about 1.1 trades a session, a real edge of $25 a trade becomes
separable from noise somewhere near session 60. Anything claimed before then is a guess
wearing a decimal point.
The bar an observation has to clear is deliberately high, because the failure mode here is
not missing a good idea — it is adopting a bad one that eleven sessions happened to flatter.
That has already happened once on this page, and the note above says where.