# Taking the other side of options block trades loses after costs

> After hedging, the dealer on large negotiated crypto options trades made +0.68% of premium on Bitcoin and -0.20% on Ethereum in the final year, against costs of about 3.5%; both lost 2.80% net.

E837 · Verdict: KILL · Cause: Below costs · Market: Crypto & DeFi · Mechanism: Microstructure · Horizon: 1h-24h · Tested: October 2026

Source: https://alphanotfound.eu/experiments/E837-options-block-trade-maker

Large options trades are often negotiated privately between a fund and a dealer, away from the public order book. The belief is that the dealer on the other side earns a premium for size and immediacy. We replayed two years of such trades in Bitcoin and Ethereum options from the dealer's side, hedged the price risk with the underlying and measured the result after 24 hours. At entry the dealer got a price 0.91% of premium better than the mark in the final year, but after option fees and hedging costs Bitcoin lost 2.80% of premium (interval -3.67% to -2.05%), against +1.0% required.

## Key numbers

- **Gross after hedging:** +0.68% / -0.20% (of premium, Bitcoin / Ethereum, final-year holdout, events with a 24-hour exit trade)
- **Fees and hedging cost:** about 3.5% (of premium; gross minus net, Bitcoin)
- **Net after costs:** -2.80% (of premium, Bitcoin and Ethereum holdout; Bitcoin interval -3.67% to -2.05%)
- **Price improvement at entry:** +0.91% (Bitcoin holdout, versus the mark, before hedging and costs; not profit)

## The idea being tested

**Belief:** A dealer who takes the other side of a large, privately negotiated options trade is paid more than the fair value, and that premium is profit once the price risk is hedged.

**Why people believe it:** Funds, miners and structured-product desks trade in size and want certainty of execution. Public order books cannot absorb that size at the quoted price, so people argue the buyer or seller pays extra to a dealer who will take the whole trade at once.

**Why it would persist:** The flow is negotiated with a small group of dealers and never shown on the public book, so ordinary market makers and slow traders do not see it.

## How it was tested

We took every block trade in Bitcoin and Ethereum options over two years and imagined being the dealer on the other side of each one. We priced the trade against the exchange's own fair-value mark, hedged the price exposure with the underlying, waited 1 hour and 24 hours, and charged the exchange's option fees on entry and exit plus a cost for each hedge trade. Trades that were not negotiated blocks (ordinary screen trades) went through the same calculation as a control.

**Pre-registered:** yes

**Data:** Options trades with price, mark price, implied volatility, index price, side, size, and the venue's block-trade and multi-leg indicators (public trade history of the main crypto options venue; per trade; 2024-10..2026-10)

**Universe:** All block-trade legs in Bitcoin and Ethereum options. Legs sharing one timestamp form one event, so a multi-leg block is one event. Bitcoin: 84,120 events.

**Signal:** Maker sign for each leg is minus the taker's side. Instant edge E0 = maker sign x (mark - price) x index x size. Hedged markout M_H = maker sign x size x [(mark_H x index_H - price x index_t) - delta x (index_H - index_t)] for H = 1 and 24 hours, with Black-Scholes delta at the trade's implied volatility, zero interest rate and the standard expiry time.

**Execution:** Enter at the block price, hedge with the underlying at that delta, mark out at the nearest trade in the same instrument within 30 minutes of the exit time. A leg with no such trade, or whose option expires before the exit time, is dropped, and an event needs all its legs. For Bitcoin 27,268 of the 84,120 events have a 24-hour markout, which is 30-35% of events.

**Costs:** Option fee of 0.03% of the underlying value, capped at 12.5% of the leg's premium, paid on entry and again on exit. Hedge cost of 2 bp of |delta| x underlying value on each of the two hedge trades.

**Controls:** Validity gate: on ordinary screen trades, buyers must print at or above the mark on average and sellers at or below it; otherwise the side flag is misread and the test is void. It passed: buyers paid above the mark in 56% of Bitcoin trades and 63% of Ethereum trades, sellers received less in 65% and 70%. Negative control: ordinary screen trades of the same instruments, same maker rule, same markout.

**Sample split:** Discovery 2024-10 to 2025-09, holdout 2025-10 to 2026-10, decided on the holdout. Interval: bootstrap over calendar weeks, premium-weighted.

**Pass criterion:** Holdout net 24-hour result of at least +1.0% of premium, week-clustered interval lower bound above 0, median event above 0, instant edge above 0, and both Bitcoin and Ethereum above 0 in the holdout.

**Kill criterion:** Holdout net 24-hour result at or below 0, or a week-clustered interval that includes 0. Also a kill if the instant edge is positive but the 24-hour result is not (informed takers).

**Power:** Not computed in advance. The week-clustered intervals are 0.97 and 1.62 points wide on Bitcoin (discovery and holdout).

## Result

**Why it was rejected:** The holdout net 24-hour result was -2.80% of premium on both Bitcoin and Ethereum, so the pre-registered condition for a kill (net at or below 0) fired and the +1.0% pass bar was missed.

**Observations:** —

**Net mean:** -2.8%

**95% interval:** -3.67% to -2.05%

**Median:** —

**Worst:** —

- Bitcoin holdout: the instant edge over the mark was +0.91% of premium (discovery +1.17%). After hedging and 24 hours the gross result was +0.68% (discovery +1.02%). After fees and hedge costs the net was -2.80% (discovery -2.49%, interval -2.99% to -2.02%).
- Ethereum holdout: instant edge +0.72% (discovery +0.83%), gross 24-hour result -0.20% (discovery +0.48%), net -2.80% (discovery -1.45%).
- Gross minus net, a figure derived from these results, is 3.51 points of premium for Bitcoin discovery and 3.48 for the Bitcoin holdout: the entry fee, exit fee and hedge trades together cost about 3.5% of premium, several times the edge paid at entry.
- Single-leg trades, combination trades and the three size terciles were all net negative in the reported cuts (5 groups).
- The screen-trade control showed an instant edge of +0.3% to +0.6%, a gross 24-hour result of about +0.4% (gross, not net) and a net of about -3%. The block instant edge was about 1% of premium against 0.3% to 0.6% on screen, and both lose after costs.
- For Bitcoin the gross hedged result was positive on the subset with a 24-hour exit trade (+0.68% holdout, +1.02% discovery); for Ethereum it was +0.48% in discovery and -0.20% in the holdout.
- Net was negative with a positive instant edge, so the pre-registered kill clause labelled informed takers also formally fired. The label is not a finding, because the test does not identify who the takers were; costs of about 3.5% are what made the result fail, against a gross hedged result that was itself small (+0.68% Bitcoin, -0.20% Ethereum).
- Only 27,268 of 84,120 Bitcoin events have a same-instrument trade near the 24-hour exit, so every 24-hour figure rests on 30-35% of events.

### A 1% edge becomes a 3% loss (% of premium)

Bitcoin, from the instant edge over the mark (+1.17% discovery, +0.91% holdout) to the gross hedged 24-hour result (+1.02%, +0.68%) to the net after fees and hedging (-2.49%, -2.80%). Instant edge covers all events; gross and net cover only events with a 24-hour exit trade.

|   | Discovery | Holdout |
| --- | --- | --- |
| Instant edge over mark | 1.17 | 0.91 |
| Gross 24-hour, hedged | 1.02 | 0.68 |
| Net 24-hour, after costs | -2.49 | -2.8 |

### The net loss is not noise (% of premium)

Bitcoin net 24-hour result with its week-clustered interval. Neither interval reaches zero, and the holdout interval ends at -2.05%, far below the +1.0% needed.

|   | Mean | 95% low | 95% high |
| --- | --- | --- | --- |
| Discovery (2024-10 to 2025-09) | -2.49 | -2.99 | -2.02 |
| Holdout (2025-10 to 2026-10) | -2.8 | -3.67 | -2.05 |

### Ethereum is no better (% of premium)

Holdout results by stage for Bitcoin and Ethereum. Ethereum's gross hedged result is already negative (-0.20%) and both end at -2.80% net.

|   | Bitcoin | Ethereum |
| --- | --- | --- |
| Instant edge over mark | 0.91 | 0.72 |
| Gross 24-hour, hedged | 0.68 | -0.2 |
| Net 24-hour, after costs | -2.8 | -2.8 |

## Interpretation

**Why it failed (interpretation, not established):** The gross result after hedging was +0.68% of premium for Bitcoin and -0.20% for Ethereum in the holdout, while entry fee, exit fee and two hedge trades cost about 3.5%. Costs, not the counterparties, decided the result. This is an interpretation of the numbers, not a separate test.

**Lesson:** An edge measured at the mark is a price improvement, not a profit. It has to be compared with the full round trip: entry fee, exit fee and hedge trades. A hedged markout that stops at the gross figure would have read as a small positive result for Bitcoin only.

**Scope:** Covers block trades in Bitcoin and Ethereum options from 2024-10 to 2026-10 in the public trade history, a 1-hour and 24-hour markout, Black-Scholes delta hedging, the stated fee schedule, and the 30-35% of events with a usable exit trade.

## Reproducibility

**Steps:** 1) Download the public options trade history for Bitcoin and Ethereum with price, mark, implied volatility, index price, side, size, and the venue's block-trade and multi-leg indicators. 2) Group block legs by timestamp into events; maker sign = minus the taker side. 3) Compute the instant edge at the mark, then the 1-hour and 24-hour markouts using Black-Scholes delta at the trade's implied volatility (zero rate, the standard expiry time taken from the instrument name) and the nearest same-instrument trade within 30 minutes of the exit time. 4) Subtract the option fee (0.03% of the underlying, capped at 12.5% of premium) on entry and exit and 2 bp per hedge trade. 5) Premium-weight, bootstrap by calendar week, split by currency, single versus combination, size tercile and holdout, and run the same pipeline on ordinary screen trades. Check the side flag first. Expiry timestamps parsed from instrument names must be in the same unit as trade timestamps (seconds versus milliseconds), or every leg looks expired.

**Grade:** COMPLETE

---
AI-generated research, checked but not peer-reviewed. Historical, experimental research, not investment advice. Dataset Terms of Use: https://alphanotfound.eu/terms