Pinned Tweet
If I were to start swing trading from scratch, I’d:
▫️Study the basics from qullamaggie.com/
▫️Review @jfsrevg’s pinned posts. All essential tools are there, shared freely by Jeff. Low noise, high-value resources.
Test, tweak, personalize...there is no right or wrong. All these are just there available for all to learn and use. Trading community should have more people like that.
He didn't even pay me to post this. 😄
Grateful for the years of independence, I'll be closing out this brokerage account. I'm thrilled to be embarking on my journey in the realm of fund management with acceptance of a direct closed ended investment from a global macro strategic fund into my one-person LLC. As the main PM with a minority stake going forward, I'll continue to share my weekly diary if time and jurisdiction permits.
For those keen on making their mark in the professional space, show your passion, be thoughtful and honest with your struggle. Leverage your Twitter presence to enhance visibility on LinkedIn as well, and consider joining USIC if you are confident of your skills. You never know whose attention you might catch.
Below are some Twitter posts and threads from the past which I hope could aid you on your trading journey. Ultimately, passion is the main ingredient to grow your proficiency in trading.
How to lose all users who are not necessarily into "western values" in one single post.
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values.
Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line.
After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs.
The values in an open model are one fine-tune away.
Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
1
195
Thank you @DeepSeekHarness . The harness is good.
and also thanks to @relakkesyang for his amazing multi agent orchestration plugin. I got everything prepared with a supervisor agent and it dispatch coding, review..with subagents automatically. Good time saving for me.
1
2
280
The DSH agent teams plugin from @relakkesyang is here. I am still learning and testing it, so far so good. it support multi model as well. Good job.
github.com/NanmiCoder/dsh-ag…
1
90
The result of midterm could be good for Elon amd his companies, he can go back to engineering and maybe keep his mind more focused.
Tesla's inflated valuation ran out of hype five years ago and it's been circling the drain ever since.
Five years ago, you could have put $150k into $TSLA or an index fund, or stuffed it under your mattress.
After inflation, that $150k is now worth:
S&P 500: $193k
Mattress: $124k
Tesla: $104k
Grandma's S&P 500 index fund is up almost 29% after inflation. Even grandpa's mattress beat @ElonMusk.
In an AI-fueled bull market, $TSLA lost to Walmart, Costco, Philip Morris, Exxon, the New York Times, and literal garbage collectors.
Musk cashed out ~$40 billion of Tesla stock. His shareholders were left holding the bag: a stock down 15% in five years, a shrinking and aging lineup, 2% global market share, "Full Self-Driving" that still needs a babysitter, and robots that can barely stand up.
Tesla has been dead money for five years.
1
196
XAI merges subs into one to offer less, Claude rebrands offers and offer less, OpenAI merging Chat and Work into same quota allowance ...
They are all giving less, I am already using DeepSeek and am considering migration everything there.
Deepseek Flash could do all the jobs.
Replying to @koltregaskes
1
294
Start with a prompt like this example, ask AI to fine tune it and include all criteria to assess. Mention the traders who fits your style...then ship the whole thing to an agent asking for full report and don't stop until it finds something. Great learning and only cost a few $
I used AI at max reasoning level to research every info possible on a few traders I like (ex: Qullamaggie, William, Stamatoudis, Flanders...)
then design, test, validate a viable strategy.
I am impressed by the quality of the research.
A structure you are use and try.
3
302
I used AI at max reasoning level to research every info possible on a few traders I like (ex: Qullamaggie, William, Stamatoudis, Flanders...)
then design, test, validate a viable strategy.
I am impressed by the quality of the research.
A structure you are use and try.
Interesting partitioning visually. Then you can backtest entries and portfolio growth on each partition and decide how to adjust and what to avoid.
A quick look at the new Leaders Map inside the Lab Report.
This is basically my process visualized.
First, define the universe. Then identify the strongest RS leaders. After that, the job isn't to chase them—it’s to wait for them to come back into the area where I actually want to do business.
For me, that means:
• Daily price sitting roughly -0.5 ATR to +1.0 ATR from the 21EMA
• Weekly structure still healthy
• Not already too extended on the higher timeframe
$SNDK $TEAM $RNG $MU $TWLO $W $CAKE $SIG $SNOW
The rest of the map shows where every leader sits in that process: in the buy area, above it, stretched on both timeframes, unwinding, etc.
Right now, only a small group of leaders are actually back in my preferred entry zone, while many others remain extended.
That’s exactly the context I want.
Leadership tells me what I want to own. Location tells me when I want to own it.
One of the core pieces of the new Lab Report in TradersLab.
I just finished massive tests on 15 years of data, including rules for stock admission, comparing holding time, sectors / themes alignment.
The result is not so exciting. 😅 My biggest lesson is : "can I position myself on the 1/3 stocks that performs the best". That's it.
📈Buying Stocks: A Pragmatic Alternative
The following outlines an alternative approach to buying stocks for active traders.
The process is:
1) Decide what a "unit" is, with the unit defined either as "risk" (such as 0.25% to 0.5% of the portfolio) or a fixed amount/percentage such as 1%.
2) At the close, you buy 1 "unit" of every qualifying stock. On Thursday evening at the close, I bought 22 "units" in 22 stocks:
nitter.gq/SteveDJacobs/status/21…
If I missed buying any because the limit order went unfilled or it qualified at the very end of the session, it would be eligible for purchase at the open the next day if the candle was green, using an ORB setup.
3) The following session (Friday in this example), the stocks that open lower (red candle) can be sold or managed using the predetermined stop-loss.
As Dan Zanger says, "Winning horses don't back up into the gate", and the goal is to find and back winning horses.
At the end of that session (Friday), the process is repeated, buying the new units that are eligible.
The process repeats, scaling into winners with additional units and selling the losers. Once a position is full size, it is trailed using the MA20/21, looking for a close below.
Should there be too many opportunities for the available capital, the stocks are chosen based on industry strength (RS) and strike zone.
The goal is not to have "lots of positions" but to take starter positions in every name that qualifies as an eligible purchase and let the market feedback tell you which stocks are winners and which are not. At the time of purchase, you have zero market feedback.
nitter.gq/SteveDJacobs/status/20…
You could define a ladder upfront. For example, 1% > 4% > 7% > 10% or 2.5% > 5% > 7.5% > 10%.
This would then take you a minimum of 4 days to get a "full position", with potentially 40+ starter positions.
The goal, however, is to get positioned in 10 market leaders by letting the market tell you which stocks become those leaders.
If you never took a starter position, all too often the stock drops off your scan and you look back at something like $TXG and say "Man! How did I miss that? ... I had that on my scan at $25 at the end of May." ... but you were too busy trying $RKLB or getting ready for $SPCX because your "gut" told you space was the theme. Even though $TXG qualified, without a starter position it was simply forgotten.
(I missed $TXG scanning myself ... my bot "Bumblebee" didn't, which is one of the reasons I am moving more and more towards full automation)
Counterarguments to this approach:
1) You cannot possibly watch/manage 50+ stocks at one time!
You can configure a TradingView table to track 40 stocks in real-time per indicator. The indicator applied twice gives you 80 stocks in real-time with Buy/Add/Reduce/Sell instructions.
2) Entering so many orders would take ages!
Maybe if you were only using your phone, but you simply schedule backward from the close based on your timing. Manually, I can enter around 4-5 orders per minute (you can set up a spreadsheet or a table in TradingView to calculate the quantity based on risk or allocation), so you can complete around 20 orders in the last 5 minutes into the close.
Over time, you can build automation to perform this process.
3) You cannot watch 50+ stock charts!
I don't need to watch the stock charts. I know where I want to buy a stock and how much (quantity) is a unit. I already know where I am getting out (stop-loss) before getting in, so I can set the stop-loss immediately after purchase, or after the close as a 60-day GTC order.
4) Aren't all the stocks simply the same theme? (Semis, so just buy $SOXL?)
At present, it is easier to simply buy ETFs since the market is heavily focused on a single best-performing sector. However, this is not true historically, across all markets and market conditions.
An ETF is simply a pre-defined basket of stocks with different weightings, whereas here we are trying to have the market get us into the best names within that sector.
This approach is not for everyone, and it is unlikely to win the USIC. Many retail traders have read about and sworn by a 2-3 stock sniper approach. My preference is to fish with a net, keep the big fish and throw back the "tiddlers" 🎣
The 100 box problem may also add clarity to what I am trying to convey.
nitter.gq/SteveDJacobs/status/21…
1
3
293
Always good to see alternatives.
I might be interested when @openmarket_xyz have: android app, custom scripts I code, screener based on my personal custom codes.
What is your plan to compete with pinescript?
CME options are live on OpenMarket.
• Chains for ES, gold, crude, copper, FX and more
• Open interest, volume and greeks by strike
• Options flow: every print, size and premium
1
230
Good and Important 👇 Don't even need to trust, run it yourself.
📈Backtesting Stocks - ONE filter to rule them all 🌟
During a recent backtesting, one of the most interesting findings was a prerequisite that universally improved the results of every system/strategy indicator - U&R, flat base breakouts, pullback to ascending moving averages etc.
Regardless of your strategy, if you restrict yourself to the top 50% of industry groups (by relative strength) - you results improve in EVERY year when compared to the same strategy without the industry constraint.
Further, if you restrict yourself to the lowest 50% of industry groups, your results will be worse than the baseline in EVERY year.
To practically apply this rule, there are 149 industry groups in @finviz_com . On a weekly and monthly basis, restrict yourself to the top 74 industry groups without exception and before anything else.
finviz.com/groups?g=industry…
That's it - no. If this then that, no complex formula.... but after 1000's of tests across 300+ unique strategies, this was undefeated.
As @RealSimpleAriel likes to say "stick to leading stocks in leading groups".... the screenshot below is the scientific, quantified proof to "stick to leading groups" ... 16 years EVERY strategy undefeated ... top 50% of industries beats the baseline (all industries) and the bottom 50% ... and it is monotonic.
2
247
Must read if you do backtesting. One of the most important filters many traders praise
📈Backtesting Stocks - ONE filter to rule them all 🌟
During a recent backtesting, one of the most interesting findings was a prerequisite that universally improved the results of every system/strategy indicator - U&R, flat base breakouts, pullback to ascending moving averages etc.
Regardless of your strategy, if you restrict yourself to the top 50% of industry groups (by relative strength) - you results improve in EVERY year when compared to the same strategy without the industry constraint.
Further, if you restrict yourself to the lowest 50% of industry groups, your results will be worse than the baseline in EVERY year.
To practically apply this rule, there are 149 industry groups in @finviz_com . On a weekly and monthly basis, restrict yourself to the top 74 industry groups without exception and before anything else.
finviz.com/groups?g=industry…
That's it - no. If this then that, no complex formula.... but after 1000's of tests across 300+ unique strategies, this was undefeated.
As @RealSimpleAriel likes to say "stick to leading stocks in leading groups".... the screenshot below is the scientific, quantified proof to "stick to leading groups" ... 16 years EVERY strategy undefeated ... top 50% of industries beats the baseline (all industries) and the bottom 50% ... and it is monotonic.
1
214
Some basics that must be covered if you do serious tests on hypothesis.
📈Backtesting With Claude - The S&P500 + Nasdaq-100 Constituents 📉
In the last backtest:
nitter.gq/SteveDJacobs/status/20…
In this backtest, we consider stocks that are currently part of the S&P500 ( $SPY) , DJIA and Nasdaq-100 ( $QQQ) only. Combined, there are 517 stocks. These are among the most widely followed and traded in the world.
Known Issues that have been "Improved"
There were some known backtest issues.
nitter.gq/SteveDJacobs/status/20…
To address some of these, the following changes have been made:
1) Minimum Market Cap at Entry
A new parameter "Minimum Market Cap" (at entry) has been added based on the formula:
Current Market Cap × (Entry Price / Current Price)
It gives an estimated market cap at entry. It is not perfect but it is better than nothing.
2) Minimum Daily Dollar Volume
A new parameter has been added as a threshold for the 10-day average daily dollar volume (D$V) at the time of entry to ensure there was adequate liquidity when the trade was entered.
For example, use 10-day D$V of $50M+.
3) Price Floor
A minimum $ share price at the time of entry as been added to ensure the stock is institutional.
Universe & Trade Eligibility
The current 517 unique constituents of S&P500 and Nadaq-100 are taken and backtested using the algorithm of the prior backtest with the improvements/constraints that:
1) Market cap at time of entry is $1B+
2) 10-Avg Daily $ Volume at entry >= $50M
3) Share Price at time of entry >= $10
The above enforce that the companies were institutional grade at the time of purchase but does not address that some of these stocks have been included in the index over time. Point-in-time data is needed for that (which will be performed later).
General Observations:
- From the 1st Jan 2020 until 25th Mar 2026, the opportunity set was 1,425 potential trades across 487 stocks.
- In the prior test, 1-2% allocation was found to be optimal due to the sheer number of potential trades. You need to ensure you were adequately diversified for the outsized winners to be part of the portfolio.
With far less opportunities, larger position sizes must be taken to achieve superior performance.
- 50% (2 stocks) gave the best returns however, few would have the appetite to stomach that volatility. Fully Kelly was 35.5%
- 25% per position (4 positions) gave a CAGR of 81.9%, while the max drawdown was 27.8%. If the portoflio only contains $SPY and $QQQ stocks during 2022, you are going to experience some drawdown.
- The strategy worked best on the Technology Sector and was terrible on Real Estate ... to the point that Real Estate should be omitted altogether.
- Even with 4 positions, the average capital utilisation was only 73.8%
In summary, the strategy outperformed the S&P500 and Nasdaq-100 indices over the 6+ years tested.
The addition of the entry thresholds have removed a lot of the then-tiny stocks that grew to much larger companies which in turn have significantly reduced the CAGR and also increased the max drawdown.
In the next posts, I will outlined the results of applying these entry thresholds to the original test dataset.
Hope this was interesting ... if you made it this far, thanks! 🙏
1
208
Was doing some backtest.
For now, raw test across all NYSE stocks meeting some basic criteria ex: >$1b
- 1st one is only my own breakout criteria
- 2nd is added some MA alignment
- 3rd is like 2, but stocks are in bullish sector
Surprising results!
3
2
249



























