Has a fee structure (0% maker fees 0%-0.03% taker fees) that makes it easy to wash trade
Are Bullish's founders running back the same (alleged) wash trading playbook now that Wall Street and retail are blindly pushing the chips to any and every IPO?
I wanted to use AI to deconstruct the financials to see whether we can find any evidence of wash trading on Bullish (and hence inflating KPIs). Which if so could be a key piece in a short thesis!
I saw this fee structure on Bullish's website and my first reaction was not how cheap it is to trade here, but rather how this makes wash trading easy (a common practice in crypto)
Here's the plan:
Use ChatGPT to generate a proper prompt about our hypothesis
Run the expanded prompt on ChatGPT alongside F-1 filing document RAG
The Prompt
Prompt 1: Crafting the Research Prompt
You are a top tier forensic accountant. I need your help to generate a prompt for ChatGPT to analyze the accounting quality for Bullish $BLSH (they report under IFRS), with a particular focus on how 'adjusted transaction revenue' is adjusted for. I am highly suspicious the company has a lot of wash trading and much of the revenue being reported on the financials is constructed in a way that conceals true revenue. I want to understand especially revenue recognition and precisely how revenue is adjusted. I also want to understand how this deconstructed revenue flows through the balance sheet and cash flow statement, and if we can find any red flags on the cash flow statement. Please include anything else I may have missed here. Prioritize primary sources (I found one filing source here: https://www.sec.gov/Archives/edgar/data/1872195/000110465925073371/tm2421409-19_f1a.htm?utm_source=chatgpt.com)
Our prompt here focuses on (1) How transaction revenues are adjusted (2) Finding any wash trading indicators on financials (3) How adjusted revenue flows through financials (4) Any disclosures hidden in the offering docs (5) Looking for other red flags in disclosures like related party transactions or rapid growth in non cash items in BS / CFS
Here's a summary of the red flags that ChatGPT found from our prompt:
Bullish's "Adjusted Transaction Revenue" (ATR) is a self-defined, non standard metric.
ATR blends actual customer fees with market fluctuations of Bitcoin / crypto inventory (this makes it VERY hard to distinguish how much is real revenue from real customers vs "revenue" from market movements) .... it's possible to get negative revenue under their definitions
Filing documents show wash trading signals:
(1) Bullish's own risk disclosures explicitly acknowledge the platform could be used for market manipulation, including wash trades
(2) bloated gross notional value that doesn't generate much cash
(3) huge customer concentration which is a classic wash trading red flag
(4) Bullish's own subsidiary BTH acts as a liquidity provider
Bullish reports high transaction revenue yet operating cash flow is negative (How do you lose money in a crypto bull market? My only explanation is most of the "revenue" is tied to appreciation of balance sheet assets)
Complex related party loans from entities tied to block.one which adds P&L volatility (and obscures core P&L)
I'm in the process of building a proper view on this company and if I will short the stock. I'll be posting a full article at some point.
It's amazing how much additional leverage AI gives us for combing through documents - the above analysis took me about a minute of prompting that otherwise would've taken hours manually. I'm fully convinced that over the next year or two, having top decile prompt engineering capabilities is the #1 skill that will separate good analysts from great analysts.