Vibe Trading

Price found some support at a confluence level, the Call Wall / GX1 and also session open. A bounce or a reversal?
I haven't had a signal fire yet at this level

That is a full ATR rotation, looking forward to whether we go to the next line above, or back to the RTH open

So V2 of the order flow charting software has been built.
It now uses Tradingview charts instead of python to display the data - live feed from Topstep
I have basically made Order Flow indicators that the TopstepX doesn't. The fact that it took me several hours makes me think it wont be long before this type of analysis will come as standard. The next stage is to query the API and pay for more DOM, and see if I can build a heatmap of passive limit orders that sit in the book, waiting for these aggressive buyers and sellers. Though that is for the future, when I have a better understanding of what these aggressive orders are doing.

Done for the day - locked out

Visual Training


Previous Analysis to the VWAP was a retracement not a reversal

There was a moment at least when there were some trapped shorts at the GX1 level. Took a quick scalp on the bounce.

I'm not trading as I have locked myself out, so just visual training

This is very close to the problem I’ve been thinking about.
Vibe coding makes it much easier to generate and test ideas, but the hard part is still moving from “this worked historically” to “this can actually be executed under realistic conditions.”
For me, the missing layer is live-like practice / execution validation: entries, exits, sizing, costs, slippage, drawdown, and review. Especially with prop accounts, the strategy has to survive the rules, not just the chart.
Curious to follow how your Topstep XFA forward test develops.

The honest answer for the XFA is that without manual intervention, the account wouldn't have lasted the month of June. A lot of what worked over the last 9-years was undone by the wide ranges that we saw last month. The volatility to be able to test either side of those ranges nearly killed the account because the setups were designed for what is essentially a less volatile market, even though all of the strategies performed +ve during the madness of COVID. Taking the learning points from the recent market regime, I have now implemented a Preservation mode, which turns off everything that is not currently working as the back testing indicated it should, and I have also added a limit to the amount of drawdown potential. So if the backtest says the tail events were 200 ticks of risk, if the current regime is 201 ticks of risk on the day, we skip the trade. If the current trade says 199 ticks, this is within the risk parameters tested for, and we go ahead with the trade.

I have faith in the strategies, it is more about risk management and surviving regimes that were not tested during the last 10-years. Obviously, as we walk forward, we have this new data, and we can work out what would have worked instead, and we're adding new ideas all of the time.

In terms of the backtesting, I trained the data on a portion, validated it in OOS and then did a walk-forward analysis before putting any of this live. If we get a repeat regime like anywhere between 2016 and 2025, all will be well in terms of the strategies, which should look after themselves. I have also looked for inefficiencies in the order flow, and I am looking at longer-term structure. The work is not done, and discretionary trading is more profitable than systematic trading (currently). I am looking to switch that around in the coming weeks/months/years. Getting away from the limitations and costs of a funded trading account, and into my own capital is the key to really pushing forward.

V3 of my trading analysis landed last night. I now have zero excuses to fail

Here is some practical advice for more investigation:

1. Location — price tagged a known level of interest, VAH (30,253.03), a level shown to draw more engagement than a random price.
2. Candle shape — the bar printed as a doji: small real body relative to its range, meaning price stalled and made no net directional progress.
3. Big-order marker in the wick, not the body — a flagged large order sat at the tip of the upper wick rather than inside the body, meaning it was rejected/absorbed rather than followed through.
4. Delta/effort-result divergence — the bar's cumulative delta was negative (-193) even though the candle itself closed green: the order flow disagreed with the candle color a full bar before price confirmed it.
5. Diagonal/stacked sell imbalance at the exact extreme — 131 sell at 30286 vs 35 buy at 30287 (3.74x, clears the 3x threshold): aggressive selling directly overwhelmed the buying attempt right where price topped out.
6. Follow-through confirmation — the next bar (10:05) printed a large red candle with an even more negative delta (-618), confirming the rejection actually materialized.

1. don't fade open drives (verified, 72-77%),
2. don't trade naked levels,
3. always have an invalidation,
4. size only with confirmation.

Subtracting those behaviours moves a typical retail trader from losing fast to losing slowly. That's genuine value, but arithmetic is against stopping there: futures costs are the house edge, and our repeated findings across every corpus is that OHLCV-visible effects are sub-cost.