[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
[@quant_arb](/creator/twitter/quant_arb)
"Stacked data: This may seem fairly trivial but actually streamlines the process for alphas where the data frequency is low enough that this is possible. Normally you would do: SYMBOLS = . tools . data_loader( start_date="2023-01-01" end_date="2025-01-01" data_type="trade_bars" frequency="1h" symbols=SYMBOLS ) and it would behind the scenes look like: dfs = for symbol in symbols: for date in pd . date_range(start end): load in the data append to dfs return pd concat(dfs) This is probably one of the most common tools to use but it's actually quite slow. It takes about XX mins to load in hourly"
[X Link](https://x.com/quant_arb/status/1978418871951311042) [@quant_arb](/creator/x/quant_arb) 2025-10-15T11:13Z 64.3K followers, 8360 engagements
"Always ensure your .resample() is: .resample("X" label="right" closed="right") By default it's left so the start of the period is the timestamp (instead of the end of it) Easy way to produce lookahead is by forgetting this. Most LLMs will not add this"
[X Link](https://x.com/quant_arb/status/1978848822575931649) [@quant_arb](/creator/x/quant_arb) 2025-10-16T15:41Z 64.3K followers, 19.9K engagements
"Quant Roadmap 2024/25 Ultimate Edition 🚨 Containing: - Hundreds of resources - GBs of free data - Guidance on strategies - Much more link below"
[X Link](https://x.com/quant_arb/status/1811884056789045297) [@quant_arb](/creator/x/quant_arb) 2024-07-12T22:03Z 64.3K followers, 307.1K engagements
"Stablecoin markets are piss easy to MM if you want some easy market making experience. Go MM stablecoin to stablecoin markets. Super non-toxic can literally do it manually. I made money doing it manually at one point"
[X Link](https://x.com/quant_arb/status/1973500088082374736) [@quant_arb](/creator/x/quant_arb) 2025-10-01T21:27Z 64.3K followers, 77.9K engagements
"I even make this mistake sometimes. You can spot this kind of lookahead in the PnL curve super easily though. It'll look like 20-50 Sharpe"
[X Link](https://x.com/quant_arb/status/1978850185825374598) [@quant_arb](/creator/x/quant_arb) 2025-10-16T15:47Z 64.3K followers, 8614 engagements
"Hetzner SX line is probably the best way to do serious QR on a budget. 300TB for $XXX a month. Can build real HFT backtesting infra in Hetzner for dirt cheap. Generally useful for side projects too"
[X Link](https://x.com/quant_arb/status/1980052369825026154) [@quant_arb](/creator/x/quant_arb) 2025-10-19T23:24Z 64.3K followers, 8126 engagements
"RAM wise can always chunk your data because 250GB vs 1TB or RAM increases cost a lot but storage is the real hurdle thats hard to get around. STUPID EXPENSIVE ON AWS"
[X Link](https://x.com/quant_arb/status/1980052653427036573) [@quant_arb](/creator/x/quant_arb) 2025-10-19T23:25Z 64.3K followers, 2692 engagements
"@cryptohound666 Just for research"
[X Link](https://x.com/quant_arb/status/1980100971239014537) [@quant_arb](/creator/x/quant_arb) 2025-10-20T02:37Z 64.3K followers, XXX engagements
[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]
@quant_arb
"Stacked data: This may seem fairly trivial but actually streamlines the process for alphas where the data frequency is low enough that this is possible. Normally you would do: SYMBOLS = . tools . data_loader( start_date="2023-01-01" end_date="2025-01-01" data_type="trade_bars" frequency="1h" symbols=SYMBOLS ) and it would behind the scenes look like: dfs = for symbol in symbols: for date in pd . date_range(start end): load in the data append to dfs return pd concat(dfs) This is probably one of the most common tools to use but it's actually quite slow. It takes about XX mins to load in hourly"
X Link @quant_arb 2025-10-15T11:13Z 64.3K followers, 8360 engagements
"Always ensure your .resample() is: .resample("X" label="right" closed="right") By default it's left so the start of the period is the timestamp (instead of the end of it) Easy way to produce lookahead is by forgetting this. Most LLMs will not add this"
X Link @quant_arb 2025-10-16T15:41Z 64.3K followers, 19.9K engagements
"Quant Roadmap 2024/25 Ultimate Edition 🚨 Containing: - Hundreds of resources - GBs of free data - Guidance on strategies - Much more link below"
X Link @quant_arb 2024-07-12T22:03Z 64.3K followers, 307.1K engagements
"Stablecoin markets are piss easy to MM if you want some easy market making experience. Go MM stablecoin to stablecoin markets. Super non-toxic can literally do it manually. I made money doing it manually at one point"
X Link @quant_arb 2025-10-01T21:27Z 64.3K followers, 77.9K engagements
"I even make this mistake sometimes. You can spot this kind of lookahead in the PnL curve super easily though. It'll look like 20-50 Sharpe"
X Link @quant_arb 2025-10-16T15:47Z 64.3K followers, 8614 engagements
"Hetzner SX line is probably the best way to do serious QR on a budget. 300TB for $XXX a month. Can build real HFT backtesting infra in Hetzner for dirt cheap. Generally useful for side projects too"
X Link @quant_arb 2025-10-19T23:24Z 64.3K followers, 8126 engagements
"RAM wise can always chunk your data because 250GB vs 1TB or RAM increases cost a lot but storage is the real hurdle thats hard to get around. STUPID EXPENSIVE ON AWS"
X Link @quant_arb 2025-10-19T23:25Z 64.3K followers, 2692 engagements
"@cryptohound666 Just for research"
X Link @quant_arb 2025-10-20T02:37Z 64.3K followers, XXX engagements
/creator/twitter::1199145208895725573/posts