Paradigm Insights | Quantitative Analysis of Paradigm BTC Option Block Trades

Back to Blogs

Published On

May 24, 2023

Over the past several years, an anonymous trader within the VIX options market earned the nick-name “50-Cent.” While it’d be interesting to see the rapper 50-Cent become a hedge-fund manager, this particular trader received the title after systematically trading massive volumes of VIX calls priced at $0.50.

At one point in 2018, 50-Cent lost nearly $200M after consecutively bleeding out losses from call options expiring worthless. However, in 2018 the VIX had a massive spike jumping from single digits to over 50. This event wiped out a lot of systematic vol sellers (aka Volmageddon), however, 50-Cent had the last laugh and made nearly $400M during this period!

Since 50-Cent received media attention, people always seem to pay attention whenever there’s a large trade of VIX calls priced at $0.50. Some investors use these trades as signals to gauge the market and subsequently trade vol given 50-Cent’s past success. 

In the case of crypto options, we can perform a similar analysis and look at historical trade data to determine whether following certain traders can provide us with valuable insights. This research piece aims to analyze the trading flow of BTC option trades on Paradigm and develop more significant insights into trader behavior.


A block trade on Deribit is any trade privately negotiated between the trader and market-maker and settled on the exchange. More prominent traders will generally use block trades to achieve better execution and reduce slippage by negotiating with a market-maker rather than relying on screen liquidity. 

One could argue that large trades still execute at screen quotes on Deribit. However, the majority of institutional option traders execute their flow using Paradigm’s RFQ venue. Focusing our analysis on Paradigm’s block trades gives us a better chance to track sophisticated and informed flow. 

The dataset used for this analysis contains over 100,000 distinct trades; therefore, looking at all of the data at once will make it hard to filter any meaningful insights. Instead, we can break up this analysis by strategy type and look at the cumulative P&L within each cluster. For this research piece, we will focus on the three strategies outlined below:

- Single Leg Options 
- Volatility Spreads
- Call and Put Spreads

Single Leg Options

Using the Paradigm historical block trade data, we can filter by the block trade ID and only look for trades that contain a single leg. Although there could be a variety of different reasons for initiating a single-legged trade, we make a few major assumptions:

- The trader is using this single-leg option to express a view on the price of BTC; therefore, there is no delta-hedging component involved 
- Regardless of the option’s maturity, the trader holds their position into maturity and settles at expiry based on Deribit’s delivery prices
- P&L is calculated using the original trade execution price and size while keeping all denominations in BTC

Using the abovementioned rules, we can calculate the theoretical P&L of every trade in the data and calculate the cumulative returns across time. Although this basic vanilla approach has considerable volatility, it ends up just about breaking even.

One method to improve returns would be to filter trades by their original trade sizes. There are a few logical reasons for this:

• As shown in the plot below, there are far more trades with smaller position sizes relative to larger trades. Specifically, trades with a notional size of 25 BTC made up nearly 50% of all trades in the dataset. By looking at smaller trades, we are forced to look at significantly more trades (many of which may be sub-par) and lose our ability to filter for any meaningful edge.

• If a trader places a larger position, they are likely either an institution or an informed player when some edge on the market. In other words, these traders are less likely to take large positions without some thoughtful analysis. We can support this theory by looking at the cumulative P&L of each trade when filtered by the trade position size. This analysis assumes a constant 1 BTC position size across all trades to keep the comparisons consistent across different sizes. For example, if we have a trade size filter of 50 BTC, then the plot will showcase the cumulative P&L of all trades greater than 50 BTC assuming a 1 BTC notional position. We can run this analysis for various position sizes to showcase that as the trade size gets larger, so does the cumulative P&L (notice how the P&L jumps up after we filter for trades greater than 25, 50, and 100 BTC notional). It’s important to note that filtering the data based on larger trade sizes reduces the number of trades used in the P&L calculations, which will reduce the robustness of the backtest.

Fortunately, this thought process is consistent with the equity curve results when we filter by trade size. Similar to the above, we assume a 1 BTC notional position size to more easily compare the profitability of each bucket without skewing the results based on large trades.

For trades greater than 100 BTC, the equity curve has a consistent upward trend. When analyzing the raw trade data, larger traders performed exceptionally well over the past year, trading successfully around key risk-off events. For example, the massive spike in P&L during May 2021 was attributed to large traders purchasing puts before BTC crashed from $60k to $35k. Similarly, in May 2022, when LUNA imploded, large traders bought a significant amount of puts in size to hedge against further downside. Conversely, smaller traders, such as those who traded within the 50 to 100 BTC trade size category, experienced massive losses in early 2021 due to buying OTM calls but failed to take profits before the price temporarily corrected.

Volatility Trades

We can build on this framework by focusing on vol-specific trades, which are defined as straddles or strangles. Typically these vol trades tend to be initiated by sophisticated traders. Therefore, the expectation is the P&L associated with these trades should be positive. Given there are no explicit labels in the data indicating the type of trade, these vol trades were identified using the following filtering criteria:

For each unique block trade ID:

- There must be only two options in the structure
- There must be one call and one put
- Both options must be traded in the same size and maturity
- The call and put must be traded in the same direction

Similar to the analysis from above, we can calculate the returns of holding the vol position until maturity; however, in this case, we’ll also introduce delta-hedging as this serves as a major component of P&L in any vol trade. We can test out the hedging using several methods:

- Hedge the net delta of the trade every hour regardless of the underlying price move
- Hedge the net delta of the trade after a 2.50% change in price since the previous hedge
- Do not hedge the delta at all and let the position run until expiration

At first glance, it doesn’t appear copying volatility trades is a profitable strategy regardless of the hedging technique. Therefore, further analysis was performed to segment trades by different position sizes. However, these approaches also failed to bring any meaningful results.

Most of the losses are attributable near the beginning of the bear market in early 2022 and further accelerated into Terra’s and 3AC’s collapse (during this time, many traders faced significant losses as they were short strangles and straddles which caused the put legs to end up deep ITM). The recent rise in P&L since the start of 2023 was primarily driven by traders purchasing MAR31 strangles and straddles with strikes around $20k. Although there are more sophisticated methods to hedge the net delta of a volatility position, we’ll unlikely see a significant improvement in total returns if simple strategies like these fail to bring decent results. Overall, there are traders out there who can profitably trade BTC vol. However, using this model framework fails to derive any meaningful insight from their trading patterns.

Call and Put Spreads

The last trade structure we can analyze with a delta-bias would be tracking call and put spreads. Similar to the volatility trade classification from above, we can classify call spreads and put spreads by using the following criteria:

For each unique block trade ID:

- There must be two options in the trade 
- Both options must be traded in the same size and maturity
- Both legs in the trade must be either all calls or all puts
- Both legs must trade in opposite directions (one leg has to be short, and the other has to be long)

Similar to the above analysis, we assume the spread positions are held to maturity. Furthermore, given these trades are usually placed with a bet on the underlying direction in BTC price, we won’t hedge any delta for the spreads. Although call spread buyers experienced strong gains during the bull run between 2020 to mid-2021, ever since the Fed began its hawkish policy, the performance of this strategy has been dismal. These significant losses stemmed from large institutional players making ill-timed bets on a bullish reversal in BTC. In a related manner, the P&L for put spreads also experienced a similar fate. Surprisingly, the put spreads also faced steep losses in the midst of the 2022 bear market. However, upon closer observation, most of these losses for the put spreads were attributable to large sales of put spreads with a positive net delta exposure.

As a final check, we can filter the trades by size to determine whether larger trades tend to perform better. In the case of the put spreads, filtering by trade size improved the total P&L of the strategy; however, the most profitable trades came from smaller-sized positions. This contradicts what we would have expected – larger players tend to have a greater edge.

Conversely, when filtering by the trade size for call spreads, larger trades have better returns than smaller traders, which is consistent with expectations, albeit the returns are more volatile.

Concluding Thoughts

In summary, using our predefined framework to analyze Paradigm block trades, we can conclude that individual option legs and calls-spreads traded in larger sizes offer the most valuable insights. Despite the sophisticated nature of vol traders, these vol structures failed to produce meaningful returns under our modeling approaches. One reason for poor performance is due to the overly simplistic assumptions made when modeling these vol trades. In practice, traders will likely exit positions early and have more sophisticated metrics to hedge their deltas. Lastly, it was surprising that smaller-sized put spread trades performed far better than larger-sized positions. A potential reason for this could be the limited number of put spreads traded greater than 100 BTC. With only 6% of all put spreads trading more than 100 BTC in position size, the lack of data can lead to outliers dominating the P&L.

As a whole, one key thing to note is throughout this analysis, there were several critical assumptions made in an attempt to simplify a vast amount of data into an easier-to-digest framework. It’s important to highlight that we don’t know precisely what traders would have done with their positions (i.e., assuming all positions are held into expiration is not practical). However, in the absence of perfect information, this is a decent starting ground for analysis to serve as the basis for a more sophisticated strategy. For example, future analysis can look at the change in open interest for each option and determine whether specific trades are being closed or opened. 

Overall, we are still relatively early in the development of the crypto options space. With the advent of decentralized option protocols and the CME starting to see larger volumes, future analysis will also have to incorporate the flows from these venues to accurately model the collective behavior of this market. As this space continues to grow, it’s only a matter of time before we see our own “50-Cent” of crypto options!

Acknowledgments: Shiliang Tang, Josh Lim, Sohan Sen, and Joe Kruy for their comments and feedback on this piece. 

We've recently updated our privacy policy. The updated policy can be found here. Continued use of our services constitutes acceptance of our updated policy.