Geometric Scaling
Position size doubles at each earned buffer — not linear, geometric: 1 → 2 → 4 → 8 → 16. Four doublings is the cooperative ceiling. Each step takes six uncorrelated trades' premium at the current size to fund.
$5K to $50K is log2(10) ≈ 3.32 doublings. The Compounding Engine’s job is to compress those ~3.3 doublings into the shortest operating window the architecture will safely permit. Where the parent Experiment instruments the Doubling Doctrine on a linear payoff — ES futures — the Compounding Engine instruments the same doctrine on a convex payoff: long short-dated options (0–7 DTE) on SPX and single-stock underlyings. The trigger ladder is geometric: position size doubles when a six-uncorrelated-trade profit buffer has been earned. The architecture installs the doctrine’s Earned Doubling Law in its strictest form — size advances only on won capital toward the $50K target, and reverts one step the moment that capital is returned.
Public method. Public trades. Public falsifiability. Anyone in the world can watch it.
Read the doctrine first → · for visitors arriving without context
A 10× run from $5K to $50K is ~3.3 doublings — and convexity is the instrument property that makes those doublings reachable. The Experiment runs on a linear instrument — ES futures, where payoff scales linearly with the underlying. The Compounding Engine runs on a convex instrument — long short-dated options, where payoff scales nonlinearly. On short-dated options near the money, gamma spikes: a small underlying move produces a large change in delta — meaning directional exposure accelerates as the underlying advances toward the strike. Risk is bounded at trade entry by the premium paid. Reward is bounded only by the underlying’s range across the option’s remaining lifetime. This is the instrument-level source of asymmetric payoff geometry; the protocol-level doubling architecture (§01.2) compounds it.
The risk side is set by the premium paid — known to the dollar at trade entry. The reward side is set by the delta acceleration if the underlying moves in the position’s favor. This asymmetry is the instrument property the protocol’s six-trade buffer requirement is designed to harvest.
Gamma instruments are regime-brittle. The same property that produces nonlinear acceleration in cooperative conditions (trend, range expansion, event-driven moves) produces nonlinear decay in hostile conditions (chop, fade, premium decay outpacing delta capture). The architecture’s defense is reactive, not predictive: the daily one-loss rule, the $600 weekly cap, and auto-revert (§04 Layer 03) fire on actual realized losses; the Falsifiability Gate (§05) catches edge degradation across the rolling 100-trade window. None depend on classifying the regime in advance.
Same posture as the Experiment: not under operator discretion, not modified under stress, the same on a winning week as on a losing one. Four pillars, then the earned-doubling ladder, then the risk parameters.
The four principles are the same as the linear track’s; only the unit definitions change for options. See the full Never Lose Money framing →
Structural edge on long short-dated options near the money — pricing inefficiency in defined timeframes plus the asymmetric payoff geometry above.
1R = premium paid. Each entry’s EV is the probability-weighted payoff minus the premium, recomputed live as fills close.
Starts at one option contract (one unit of premium-at-risk); doubles on each earned buffer per §01.2.
Daily max one loss. Weekly $600 cap. Auto-revert on buffer give-back. Falsifiability Gate.
The $5K unit doubles as the initial buffer. Working capital per unit is $5,000 — the entire allocation is at risk. Adherence-protected, not cushion-protected: so long as the stand-down architecture fires on schedule, the $5,000 stays out of the risk-of-ruin zone. Same drawdown-recovery arithmetic as the linear track (−50% needs +100% back).
Why the weekly cap is tighter on this track. Linear track: $1,000 weekly cap = 20% of $5K. Convex track: $600 = 12%. Gamma instruments are regime-brittle — faster stand-downs are the correct architectural response.
Base position size: 1 contract (1 unit of premium-at-risk) Buffer requirement at any size N×: 6 uncorrelated trades' premium at N× When cumulative profit ≥ buffer requirement at current size → Position size doubles to 2N× When cumulative profit falls back to the level at which the most recent doubling fired (the buffer just earned is fully given back) → Position size reverts to the previous step (N× → N/2×) Daily constraint: maximum one loss per day. Trade stops for the day on the first realized loss. Resumes next session.
Counted toward the six-trade buffer only if the trade satisfies all three:
This definition is locked at protocol level. Any change requires witness countersignature and a 48-hour cool-off.
Pre-funded for margin and ruin. Earned doubling starts on a floor.
Net profit ≥ 6 trades’ risk = the right to advance one step.
Throttles daily drawdown velocity while wins compound the buffer.
Give back the level’s buffer? Drop one step: 4× → 2× → 1×.
| Parameter | Threshold | Consequence |
|---|---|---|
| Daily max losses | 1 realized loss | Stop trading for the day. Resume next session. |
| Weekly max premium-loss per unit | $600 (12% of $5K) | Stand down for the remainder of the week. |
| Consecutive losing weeks | 2 in a row | Stand down for the remainder of the month. |
| Rolling 100-trade realized EV | $0 / trade | Falsifiability Expression Gate fires — see Section 05. |
| Buffer-lost-at-N× event | P&L returns to doubling level | Auto-revert to N/2×. Announced in Discord within 24h. |
| Revert within 14 days of prior revert | 1 + 1 within 14 days | Auto-pause + witness review + countersigned restart required. |
The $5K → $50K challenge is ~3.3 doublings; the table below is how many of those doublings each regime window is expected to deliver. Gamma instruments compound on regime presence, not on calendar time. A regime window is a continuous period during which entries are not in stand-down or revert-pause states — an outcome of the reactive architecture, not a precondition declared in advance. “As fast as possible” means: as many earned doublings as cooperative windows allow, with zero tolerance for unearned size.
| Scenario | Doublings (log2) | Notes |
|---|---|---|
| Sustained cooperative window every buffer earned | up to 4 doublings (16×) | Ceiling clipped pending live record |
| Mixed window partial earnings + sit-outs | 1 – 3 doublings (2–8×) | Friction from revert events folded in |
| Sustained hostile window stand-downs dominate | 0 – 0.5 doublings (~1–1.5×) | Capital preserved at base size; edge held in reserve |
| Falsifiability case Expression Gate fires | architecture halts | Strategy moves to research-only per §05 |
Critical compliance note. Until a multi-quarter live record exists, all trajectory figures are hypothetical and projected. The cooperative-window trajectory is clipped to ≤ 4 doublings (16×) regardless of what the math could otherwise support, pending live evidence.
Hypothetical performance has inherent limitations. It does not reflect actual trading and cannot completely account for the impact of real-market factors. The figures should not be relied upon as a prediction of actual results. Past performance is not indicative of future results. The Falsifiability Gate (§05) supersedes all trajectory logic.
◆ What a cooperative regime-window on the road to $50K might look like
The challenge is $5K → $50K, ~3.3 earned doublings, as fast as possible. This is one illustrative working-unit arc under the doubling doctrine on the convex instrument — the engine block that, run cooperatively, clears the 10× with room to spare: just ~3.3 of these doublings carries $5K past $50K. The engine compounds via earned doublings, position size growing geometrically 1 → 2 → 4 → 8 → 16 contracts as each six-uncorrelated-trade buffer is earned. Cooperative ceiling: 4 doublings (16×) on the working unit — clipped pending live record. Hard constraint: the trailing floor at HWM − 12% of the working unit, matching the weekly $600 cap. The X-axis is calendar months for visual comparability; the real driver is regime-window time — gamma instruments compound on regime presence, not on calendar.
Engine target
16× ($5K→$80K)
cooperative ceiling · clipped at 4 doublings
Doublings
4 (max)
buffer ladder · 1 → 2 → 4 → 8 → 16
Hard constraint
Peak − 12%
weekly $600 cap · tighter than linear
Position scaling
1 → 16 contracts
earned by 6 uncorrelated trades per buffer
Realistic envelope
2× — 8×
mixed-window band · 1–3 doublings
Per-trade risk
1R = premium
varies per trade · clean R unit
↑ All eight kill conditions are armed at every step
The ladder halts when any one fires — regardless of which step has been reached. These are the existing sit-out, falsifiability, fidelity, and revert conditions on the convex track; nothing new is invented for this visualization.
Position size doubles at each earned buffer — not linear, geometric: 1 → 2 → 4 → 8 → 16. Four doublings is the cooperative ceiling. Each step takes six uncorrelated trades' premium at the current size to fund.
On convex instruments, revert is not graceful step-down — it is regime-response. If the buffer at N× is given back, size reverts to N/2× immediately on the next entry. Two reverts inside 14 days triggers an auto-pause + witness review — a confirmed condition shift.
Gamma instruments are regime-brittle. The convex track runs a $600 weekly cap (12%) versus the linear track's $1,000 (20%), and a daily one-loss rule on top — faster stand-downs are the correct architectural response.
Aspirational and illustrative. Headline targets (16× engine ceiling, 4 doublings, 12-month window) are design intent, not forecasts — the cooperative scenario where every buffer fires and no revert occurs. The realistic envelope per the trajectory table is 2×–8× (mixed window, 1–3 doublings). The trailing floor at HWM − 12% mirrors the structural intent of the existing convex sit-out cadence. No live convex trades on record yet; the Convexity Experiment launches 6/30/2026. Past performance is not indicative of future results.
R definition: 1R = premium paid at entry on a single long-option position. No spread width to specify. No leg-management drift. No early-exit slippage. The R unit is what cleared the broker at the entry print. Until 80 qualified trades execute, the Expression Gate is armed but not yet binding; the trade-count threshold (80 provisional / 100 binding) is calibrated on trades, not calendar time — precisely so it cannot be gamed by extending the calendar.
Win Rate
—
of closed fills
Profit Factor
—
gross profit / gross loss
R-Expectancy
—
per unit of risk deployed
Avg risk / trade
—
premium paid per trade · 1R
Net P&L
—
cumulative · closed fills
Closed trades
—
— W / — L
Best fill
—
—
Worst fill
—
—
Every closed long-option fill, in sequence, from the published option-alerts journal. This is the model-portfolio record leading into the live experiment — educational, premium-at-risk only. Hover a point for the running total.
Three rows for institutional context. The Ekantik row is the architecture’s pitch — same instrument family, different operator, different rules. No live convex trades on record yet; the row will populate from closed fills once the live experiment opens on 6/30/2026.
| Strategy category | Edge / occurrence | Frequency / mo | Annual R (unsized) | ½ Kelly Allocation |
|---|---|---|---|---|
| Institutional Vol Trading Desks (Jane Street / SIG) | +0.05–0.15% | very high | reference | institutional-only |
| Retail 0DTE / Short-Dated Options (aggregated) | negative | high | strongly negative | 0% (negative edge) |
| Ekantik · Short-Dated Long Options pre-launch · no live record | — | — | — | — |
Sources: standard institutional vol-desk disclosures, ESMA retail-options loss-rate aggregates, FINRA consumer alerts. The Ekantik row populates from the convex-track trade log once the live experiment opens on 6/30/2026; no figures shown until a meaningful sample accumulates.
Adherence-protected, not cushion-protected. The $5K working unit (§01.1) doubles as the initial buffer; the stand-down architecture is what prevents it from reaching the risk-of-ruin zone. Same framing as the linear track →
Daily 1-loss · Weekly $600 cap · Monthly two-week stand-down. Same escalation pattern as the linear track, with tighter caps (12% vs 20% of unit) because gamma is regime-brittle.
This is Condition V of a legitimate founding thesis — named abandonment condition — operationalized in code, witnessed under countersignature, and structurally impossible for the operator to modify under stress. Structurally identical to the linear track’s gates; calibrated to the convex instrument.
Same gate as the linear track, calibrated to the convex instrument: rolling-100 realized EV ≤ $0/trade — active (provisional) from 80 trades, binding at 100. R = premium paid. Action on trigger and re-deploy gate are identical to the linear gate.
Read the convex protocol (v3 draft) → · See the full Edge Gate framing (linear track) → Doc Ref: FP-V3-2026-06 · draft, witness countersignature pending
Same three criteria as the linear track (attribution H2/H3, rule-modification integrity, daily-routine adherence ≥ 95%). The one convex-specific difference is in Criterion 03 below. See the linear-track criterion cards for full framing →
Same predisposal-log requirement as the linear track, with one addition: the scaling-rules log is bidirectional — doublings AND reverts both watched. This is the convex-track-only differentiator, because revert is primary regime-response here (§04 Layer 03), not a graceful step-down.
Binding Interpretation Rule. Expression-Gate firings during an active Fidelity T2/T3 breach are interpretation-suspended pending remediation. The architecture refuses to declare an edge falsified from data generated during a fidelity breach. See FP-V3 →
Free · Unlimited · Public Discord community
Watch every options trade as it executes. See the EV calculation update fill-by-fill. Watch the buffer ladder build, double, and revert. See the Falsifiability Gate run. No fee, no cap, no relationship required.
The Convexity Experiment runs on a separate Discord server from the Experiment. Observers wanting to watch both must join both servers — the separation keeps fills and Falsifiability events cleanly attributed to each experiment.
Concepts · Methodology · Context · Fit
If something about the experiment raises questions — about the convex instrument, the doubling architecture, the revert mechanism, or how to think about this in your own context — the operator is reachable directly.
“The best metaphor for the last decade is building aircraft autopilot software. You document failure modes before passengers board.”
— Hiren Desai · Founder & CIO · Ekantik Capital Advisors
The panels below are computed live from the model-portfolio track record published in the option-alerts journal — educational, premium-at-risk only, recalculated as each fill closes. Live-capital deployment opens 6/30/2026; until the sample reaches 80 qualified trades the Expression Gate is armed but not yet binding.
No revert events on record.
| ID | Date / Time | Underlying | Strike | DTE | Side | Entry premium | Exit premium | $ P&L | Result | R | Note |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Pre-launch. Live fills begin 6/30/2026. | |||||||||||
Each row is a single closed long-option position. Sample is pre-asymptotic until 100 qualified trades.
Each alert — entry, strike, stop, and exit — is posted to the public #option-alerts channel in real time. Model portfolio, educational only.
Questions about the Convexity Experiment, the convex instrument, the revert architecture, or how to think about this in your own context? Drop a note. The operator reads each one personally. This is a private inquiry path — not an application, not an offer acceptance, and not an expression of intent to participate in anything.