Altair
Skip to content

Trading Analysis Report: UNH

Analysis date: 2026-10-14 · Model: claude-opus-5-5 · AI-generated, not financial advice.

  • Analysis date: 2026-10-14
  • Rating: Hold
  • Generated: 2026-10-15 00:25:07
  • TradingAgents 0.6.0: anthropic, deep claude-opus-5-5, quick claude-sonnet-5-5
  • Analysts: market, sentiment, news, fundamentals; research debate rounds 5, risk debate rounds 5
  • Data vendors: core_stock_apis yfinance, technical_indicators yfinance, fundamental_data sec_edgar,yfinance, news_data yfinance, macro_data fred, prediction_markets polymarket

V. Portfolio Manager Decision

Rating: Hold

Executive Summary: Hold UNH at 395.62. Do not add here, and do not trim a position that sits within your risk ceiling, defined as the lesser of your own standard weight and 4% of the book at market, which is sized so that a 25% one-day gap costs about 1% of the book. If you are flat, wait until the Q3 release passes written criteria, then work a starter of about 1.3% on day orders at 388-391 placed after the first half hour, and add only on later-session price proof. If you are 0.5pt or more over the ceiling, sell about half the excess at the next session and the rest into 398-404 within 10 sessions, or at once on a close below 372.80 or a failed read; treat 379.60 as an alert, 372.80 as the cut-to-half line for pieces above the starter, and a weekly SuperTrend flip (now 362.97) as the cut on the rest, over a 3-6 month horizon with profit-taking at 425 and 434.

Current Price: 395.62

Price Target: 421.0

Confidence: Medium

Time Horizon: 3-6 months

IV. Risk Management Team Decision

Aggressive Analyst

Aggressive Analyst: I'm backing Hold, and I'm doing it as the aggressive voice. Hold here isn't timidity. It's the way to get the most upside per unit of pain from a stock that pays out in big, lumpy moves.

Look at what UNH has done. It went from 255.97 in March to 433.68 in July, a 69% move in four months. It slid 15.8% to 365.20 on October 1 and has since recovered about 8%. That is the return stream you're buying, and it comes with a 19.6% one-day gap in January. The trader's plan accepts that openly. A 1% loss budget against a 19.6% gap caps the full position near 5%, so the worst day is paid for in advance and you can hold through everything short of it. That's how you take a big risk without being wrecked by it.

On valuation, the conservative will say the earnings yield roughly equals the 10-year and the premium is thin. That's right: a $20.50 EPS floor over 395.62 is about 5.2%, against a 5.21% Treasury. But a Treasury coupon doesn't grow, and this earnings stream is guided to. Forward EPS of $23.14 gives a 5.85% yield, and the CFO says the low end of 13-16% growth is a reasonable 2027 starting point. Trailing free cash flow is $26.5B, about a 7.5% yield. I'll concede H1 was front-loaded, so haircut it, but even a heavy haircut doesn't leave it below the 10-year. Add a PEG near 1.0, a forward P/E of 17.1 and a beta of 0.6, and I don't see a stretched valuation.

The conservative will also point to the three target cuts after a beat. Look at where the cuts landed. Baird's 427 is about 8% above the price, and Deutsche's 438 with a Hold is about 11% above. The five-name average of 466 is nearly 18% above, and Goldman and Piper are 27-28% above. A target cut that still sits above the price looks like a discount-rate adjustment, not a bearish call. The 10-year is up 64bp since mid-July, so that's the obvious reading, though the headlines don't confirm it.

The rates backdrop is ugly, with 97% odds of no cuts and 82% odds of another hike. But it's known and priced. The stock fell from July to October 1 as yields rose. The 10-year peaked at 5.31% on October 5 and is 10bp off that, while UNH has rebounded with yields still at 5.2%. It has already absorbed much of the shock.

On technicals, I'll give the conservative real ground. The October 13 gap-up reversed fully on 12.65M shares, more than double the early-October average. The 398-404 zone has rejected price repeatedly, OBV isn't confirming, and the daily Z-score is +3.29. That's exactly why this is a Hold and not a Buy. But it isn't a sell either. The weekly Z-score is -0.18 and the monthly is +0.81, so only the daily is stretched. Part of that is a quiet-base effect: the 20-day standard deviation is about 6.0 against an ATR of 11.2. ADX has gone from 14 to 19 to 23, which looks like the early part of a trend. Trimming here means selling a stock with an up weekly SuperTrend, price above the 10, 50 and 200 averages, eight straight positive MACD histogram readings and its first RSI above 50 in a month. That gives up the convexity you were holding for.

The neutral will say wait for ADX above 25, MACD above zero, OBV turning and a close over 400. I'd point to the math. Enter at 391 with the stop at 362.97 and the average target at 466, and you risk about 28 to make about 75, roughly 2.7 to 1. Wait for the confirmation and enter near 405, and you risk about 42 to make about 61, roughly 1.45 to 1. Waiting for every box to be checked costs nearly half your reward-to-risk. The plan avoids that with a limit bid at or below 391-397, never above 400, so you get filled on a dip instead of chasing. Then it scales in: a starter third after a clean MCR read and the next third on a second confirmation. That's quick without being blind.

I'm also not leaning on the weak inputs. The StockTwits sample is one day with four unique bullish authors. The news tool's "40% upside" and implied price near $360 doesn't match the 395.62 tape, so I'm discarding it.

The Q3 medical cost ratio is the real gap. Headlines say the quarter beat and the EPS floor was raised, but no tool gives us the MCR. Operating margin slid from 8.0% to 7.1% between Q1 and Q2, and that's the bear's best fundamental fact. I won't wave it away. But reading the MCR costs a day, and being wrong about cost trend could cost another gap. The aggressive move is to read it today and be ready to act on it immediately.

Hold keeps the recovery, avoids paying up at a known supply zone, and pre-commits capital at levels where the odds favor us. That's where I come down.

Final transaction proposal: HOLD. Aggressive Analyst: Let me take back some ground first, because a few of my numbers deserved the hit they took. Then I'll show you where each of you still loses me.

The 2.7 to 1 leaned on 466, which is five headlines from a sampled feed and sits above the 453 monthly line. I'll stop quoting it. I'm also dropping the free cash flow yield and the PEG as support, since one is flattered by first-half timing and the other is built on a 42% rebound off a trough. I accept that ADX going from 14 to 23 is mostly what a 4.7% day does to the math, and without the DI lines I can't read direction from it. On sizing, I'll take the neutral's 4% ceiling and a starter near 1.3%. I won't spend credibility arguing four versus five.

Look at what's left after those concessions. Take profits in thirds at 425, 434 and 453, which are the June close, the July closing high and the monthly line, with no credit for anything above. From 391 against the 362.97 weekly stop, that blends to about 1.7 to 1, so you need to be right roughly 37% of the time to break even. That's not what I sold you before, but it's still a trade, and it uses the conservative's own prices.

The conservative says either gaps matter and the stop-based ratio is fiction, or the stop works and I'm oversizing. That's a false choice, and the neutral has it right. The ratio describes the ordinary path and the size describes the tail. Because a starter near 1.3% costs about a third of a percent of the book even on a 25% gap, I can hold a structural stop at the weekly level instead of a tight one that noise takes out. And "the next worst thing is outside the budget by definition" proves too much. It's true of every share of this stock at every weight. If it's a reason not to size to the worst gap you've seen, it's a reason not to own UNH at all, and nobody on this panel is arguing that.

I think the conservative's plan costs more than it protects. A daily close under 379.60 cuts you to half, and that level is 16 points below the close. The stock moved 17.9 points in one session today. A line less than one ordinary day away is a whipsaw generator, not a risk control. It also does nothing about the tail the conservative says is the real worry, because a January-style gap goes straight through 379.60, 372.80 and 362.97. Size protects against gaps, and the starter already pays for that. I'm with the neutral: 379.60 is an alert, and the October 13 low at 372.80 is the failure line.

The same goes for trimming below standard. Everyone agrees not to add at 395 and to sell only the excess into 398-404. Going to "a little below" standard is a bearish call, and the evidence doesn't support one. The weekly SuperTrend is up, price is above the 10, 50 and 200, and the MACD histogram has been positive eight sessions running.

I wasn't skipping the monthly SuperTrend. It sits at 453.42, 14.6% above the price, so it can't flip until most of the move is made. It's a ceiling to plan the last third around, not an entry filter. If you wait for it to turn, you're buying the July high, not a recovery.

On distribution, I'll give the conservative the pattern. July 16, September 9 and October 13 each spiked early and got sold into the close, the last on 12.65 million shares. But today did the opposite. The stock opened at its low of 382.29, bid all day, and closed in the top 6% of its range with a higher low than yesterday's 372.80. That's one session, and it hasn't touched the 398-404 zone, which is exactly why I'm not buying at market. But OBV is the weakest leg of the distribution case. About two-thirds of its slide since September 14 is October 13 alone, a session that closed down 0.2%. OBV handed all 12.65 million shares to the sellers on a day that finished flat. That's a quirk of the indicator, not a campaign.

On valuation, I agree the 5.2% yield on the floor carries no obvious premium. The neutral is right that the premium runs from nothing to about 65 basis points. I never said cheap, only not stretched. The 4.1% trailing yield rests on a window that still holds Q3 2025 at 2.59 and a Q4 near 76 cents, and both roll off. First-half GAAP EPS is already 12.94, so if the figures are comparable, the 20.50 floor needs only about 7.6 from the second half. Earnings do front-load, so I'm not calling that a forecast, but it shows the floor isn't a stretch.

I'll own that the discount-rate story for the target cuts is unproven. But the conservative's reading is a story too. Baird, Jefferies and Deutsche cut by roughly 6 to 8%, Goldman and Piper raised, and Mizuho named it a top pick. Five names from a sampled feed is not a revision trend. I'm not using it to justify buying, only to say it isn't a sell. Likewise, "priced" and "still building" are both assertions. The conservative can't build a fragility thesis on one unverified StockTwits post about rotation and then fault my sourcing. What the tape shows is a 4.7% gain on a day the broad market fell on hot inflation.

The "waiting is cheap" argument also fails. If the cost ratio is clean and 391-397 is reachable, a resting bid fills on its own. Waiting adds nothing that the order doesn't already do, and the order costs nothing. As for the event calendar of enrollment, peer prints, the FOMC and the midterms, it runs through the December 8-9 FOMC. That is the permanent condition of owning a managed-care name, not a reason to hold off until a particular date. It argues for staging, which is why the first piece is a third.

So I'd run the neutral's plan with two edits. First, adopt the cost-ratio criteria defined in advance, and if the read is clean, put the bid on the book the same day. Second, on the 404 breakout path, require the close above 404 on volume over 5.5 million but don't also demand an OBV turn. An up day on that volume lifts OBV mechanically, so you'd be counting one fact twice. Keep OBV as a condition on the retest path, where it tells you something.

We all agree on Hold. The remaining dispute is a handful of rules, and the test for each is whether it protects you from the tail or just taxes you on the normal path. Size protects the tail, and staging gets you to the upside after the evidence arrives. Cut lines inside one session's range, trimming below standard, and stacked confirmations before the first piece do neither.

Final transaction proposal: HOLD. Aggressive Analyst: I'll sign most of the neutral's ledger, because it's right. I can't call 453 a ceiling when I plan the exit and a target when I do the arithmetic. It comes out of the expected return and becomes the runner. With a sale at 425 and a trail to the 391 entry, the trade pays about 0.4 of a risk unit at the first target, about 0.9 at the second, and about 1.6 only if 453 trades. The trail also isn't the whipsaw I objected to. 379.60 sat 1.4 ATR under the close, while 391 after a 425 sale sits about three ATR under price. Distance in ATRs is what matters, so the conservative's charge of contradiction doesn't hold.

I should also say plainly what that math shows. From 391, 425 is about three ATR up and the stop is about two and a half down. On a coin-flip tape this structure is roughly break-even. It shapes the payoff but doesn't create edge, so the edge has to come from drift. The drift case is a company that raised its earnings floor again while the five-name target average slipped only 3%. That's why the first piece waits on the cost ratio and nothing else. It's the one fact that tests the drift, and a third of a 4% ceiling is a fair price for that test.

On the second piece, the neutral caught something I missed. I spent two rounds on what to require above 404 and never asked what the add pays. Buying at 406 against a 372.80 failure line risks 33 points for 19 to 28 of reward. The 386-388 retest risks 14 or 15 against the same targets, which is 2.5 to 3 to 1. So the retest is the main path. I'll keep the breakout add at half a third with a trail, and I'll defend it on cost alone. Half of a third of 4% is about two-thirds of a percent of book, and a full loss to 372.80 costs five or six basis points. That's cheap insurance against sitting flat through a gap-and-go after a clean read, the one path where a resting bid never fills. I won't call it a 2-to-1 trade. On the confirmation bar, I agree a second close above 404 plus roughly 6-7M on the first close is enough, because demanding 7.8M copies one day's volume.

On the conservative's September example, I think it shows the plan working. A 391 entry in early September touches today's 362.97 level by 37 cents intraday on October 1, but the stock never closed below 365.20, so the weekly-close rule wouldn't have fired. A 7% slide on a 1.3% starter costs about nine basis points of book. The grind explains why the starter is small and the stop structural. It doesn't argue for a tighter stop inside the range. I'm glad 372.80 only binds pieces beyond the starter, because that's where it belongs.

On peer prints, the neutral's veto-not-gate framing is right. "Wait for a couple" hangs on dates the tools told us are unverified, so it has no end date. A bad Centene or Elevance print that drags UNH down hands you the 386-388 retest or lower, which is already the main path. A veto covers that case without a postponement.

On burden of proof, the conservative says that when evidence is a wash, the burden sits on the action that adds risk. The plan already makes each tranche carry it. The starter needs the cost read, the second piece needs price proof, and the third needs MACD above zero and ADX over 25. Which action adds risk also depends on a book we can't see. For someone at standard weight, selling below it is a call against an up weekly SuperTrend. I'll also drop the "market fell on hot inflation" line, since it came from one unverified post.

I do want to use the conservative's pattern of failed spikes. The September 9 and October 13 spikes both printed at the top of the 398-404 zone. A limit order to sell any excess over standard weight there has a history of filling, so that observation tells you where to sell, not a reason to stay out.

On valuation, the neutral is right that $7.56 for the second half says 2026 is largely booked, not that $23.14 is safe for 2027. I'm not calling 17.1x cheap. I'm saying 19.3x a raised floor, with a premium over bonds somewhere between nothing and about 65 basis points, is not a sell signal. The open question is the Q4 exit rate, and that sits in the release. Write the criteria before opening it: full-year cost guide held or improved, no adverse prior-period development, and a floor raise that comes from operations.

So here's where I land. If you're over standard weight, sell the excess into 398-404. If you're at or below standard, sit and don't add at 395. If you're flat, read the release against those criteria, then bid for a starter of about 1.3% at 388-391, cancelling on a close below 382.29. Add mainly on a held retest of 386-388, or at half size on a second close above 404 with a trail. Sell at 425 and 434, trail to entry, and treat 453 as a free runner. Treat 379.60 as an alert, cut above-starter pieces to half on a close under 372.80, and review the starter on a weekly close under 362.97. Caution only gets paid when it's attached to a price, and every line above has one.

Final transaction proposal: HOLD. Aggressive Analyst: I'll sign the corrections first. The conservative and neutral are right on several points, and I'd rather lose those than defend rules that don't do what I said they do.

The cancel-below-382.29 rule was decoration. A bid at 388 to 391 fills on the way down, so by the time a close could breach 382.29 you already own it. Day orders placed after the first half hour, re-armed each morning after checking news and the peer calendar, fix that. It has a cost: today's open was the day's low, so a delay can forfeit a dip. But I'd rather miss a fill than be the liquidity on a news gap. I also accept that buying the retest blind at the shelf puts 372.80 only 1.2 to 1.3 ATR away, which is the distance we called noise a round ago. Buying after the bounce near 392 is about 2 to 1, not 3 to 1. September wasn't a test of anything, since the weekly line sat within forty cents of the low. Keying the review to the current weekly SuperTrend line, which only ratchets up, beats pinning it to 362.97. And I'm in on splitting the second piece. A held retest shows there's a bid, and a second close over 404 shows the supply can be absorbed. Those are different facts, so each half stands alone, and a gap-and-go after a clean read still earns the breakout half.

Now the pushback. On drift, I agree that stop-and-target geometry creates no edge and that revisions belong in neither column, so I withdraw my "slipped only 3%" line. But the conservative's "small negative expectation on zero drift" treats the gap tail as one-sided. You cite at least three big gap days this year but have a size and direction for only one, the January drop. Nothing in our data gives a direction for April 7 and April 21. The stock closed at 255.97 on March 27, and an insider sale printed at 355 on April 23. Both gap days sit inside a 39% run in under four weeks. I can't prove it, but I'd bet at least one was up, and we're long. The downside tail is real, and the starter is sized for it at about a quarter of a percent of book on a 19.6% gap. A zero-drift benchmark also ignores roughly $9.70 a share of annual dividends.

I'm not claiming a measurable edge. The edge is an up weekly trend, about 19x a raised floor, and a premium over bonds between nothing and 65 basis points. A starter this size is the price of owning that option. On "the stock already moved on it," the net move since the print is about 4.5% over two sessions, from 378.52 to 395.62. Price is back near the late-August and early-September closes around 398, and 8.8% below the July closing high. I won't call that under-reaction, but it isn't a stock that has run away on the news either.

The plan has also never priced the last third. Waiting for MACD above zero and ADX over 25 almost certainly means buying in the 410s. Against a 372.80 failure line that is about 37 points of risk, with 43 points to 453 and only 24 to 434. That's the flaw the neutral caught at 406, and it's worse here. So I'd stop sizing in fixed thirds and size later pieces so a loss to their failure line costs the same as the starter's loss to the weekly line, about nine basis points. At 410 that comes to roughly 1% of book instead of 1.3%. Then ride it with the ratcheting weekly line and be honest that it's a regime bet, not a 2 to 1 trade. One more note on 453: a SuperTrend line in a downtrend only moves down, so 453.42 is the highest the monthly flip level can be. I can't verify how the tool updates it intramonth, so I'm not building on it.

On peer prints, I accept veto, not gate. But define the veto before the print, for the same reason the neutral wanted the cost-ratio criteria written down first. A peer cutting full-year guidance or citing medical cost trend above plan freezes additions and skips re-arming the starter. A soft sector day alone doesn't count. Otherwise the veto becomes a way to rationalize whatever we already wanted to do.

I keep pressing on size rules rather than entry gates because of the arithmetic. A 1.3% starter that runs to 434 makes about 15 basis points. One that sits through to 362.97 loses about nine. At the starter level the whole trade is rounding error. What matters is whether the size grows when price is paying you, and whether each later piece is sized to the risk it carries. Gates that delay without adding information, like waiting on a couple of peer prints with unverified dates, cost you the part of the trade that could matter. Staging is where the convexity is, and a long stock position isn't convex without it.

Here is where I land. If you hold more than standard weight, sell the excess into 398 to 404 in a couple of clips, and don't go below standard. If you're at or below standard, sit and don't add at 395. If you're flat, read the release against the criteria written in advance: cost guide held or improved, no adverse prior-period development, and a floor raise that comes from operations. A missing or ambiguous read means no starter. Then work about 1.3% on day orders at 388 to 391, placed after the first half hour and re-armed each morning.

Add the second piece in two halves, one after a retest of 386 to 388 has held and bounced, and one after a second close over 404 on roughly 6 to 7 million shares. Add the last piece only once MACD is above zero and ADX is over 25, and size both by distance to the failure line. Sell at 425 and 434, trail to entry after the first sale, and treat 453 as a runner. Treat 379.60 as an alert, halve anything above the starter on a close under 372.80, and review the starter on a weekly close under the current weekly line.

Final transaction proposal: HOLD. Aggressive Analyst: Most of what's left I'll sign, so let me say what I'm signing and then go after the places where a rule still costs more than it protects.

Conservative, you're right that the dividend isn't a cushion. 2.46% against 3.75% cash is a deficit, and I'll stop leaning on it. You're also right that UNH sat near 398 in late August and early September and then fell about 5.5% to 376.77 in three sessions. Being back at a price isn't the same as being through it. But that supports my side of the plan. The 398-404 zone has rejected price twice this fall, so it's where the excess gets sold and where I won't buy.

Neutral, you're right that I swapped one unknown for another. My bet that one of the April gaps was up was a bet, not data. The sign of the tail is unknown, and the starter is sized for a down gap either way.

Your sizing math also checks out on my end. Holding each piece to the starter's risk of nine or ten basis points gives about 1.9% for a retest half near 392, 1.15% for a breakout half near 406, and about 1% for the last piece at 410. The half-third binds on the first two, and the total is near 3.7%, under the cap.

I take the rest of the machinery too: - Weekly line: a flip halves what's left, and a weekly trend still down a week later exits. I can't verify how this tool builds the line on a flip, so the exact wording should be checked against its output. - Holder backstop: if the excess hasn't sold by a close under 372.80, it goes then. That's a stop on the part of the position nobody sized for gaps, and it's the most useful rule added this round. - Ceiling: the lesser of your own standard weight and 4%, in both branches. - Peer veto: five trading sessions, renewed only by another adverse print. That can chain through a reporting season, and I'm fine with it. If three peers in a row cite cost trend above plan, staying out is the right answer.

Now the three places I'd still push.

First, "nothing gets added after the first sale." The reason is no chasing, and I agree with the reason. But the rule is blunter than the reason. Say you sell a third at 425, the stock gives back 9% to the 386-388 shelf, and it holds. Buying there isn't chasing. It's buying more than 30 points below your last sale, at the best ratio in the plan. The rule that does what the reason says is no adds above your last sale price. The neutral's no-catch-up rule is a separate thing, and I'm fine with it.

Second, the market-value cap. I'd apply it at the moment of an add: if the add would take you over 4% at market, you don't add. But I wouldn't force a sale because a winner drifted over the line. The planned sales at 425 and 434 already do that work, and a mechanical trim on top turns the runner into a rule violation.

Third, the implied Q4 test. It's a good idea, but the floor is "at least $20.50," so the implied Q4 is a minimum, not a forecast. With H1 GAAP EPS at 12.94, the second half is at most 7.56 if the figures line up, against 12.94 in the first half. That steep step-down is either prudence or cost build, and the arithmetic can't tell you which.

So write the criterion before opening the release. Judge Q3 against the path the old guide implied. A low implied Q4 passes only if the release or call attributes it to seasonality and prudence rather than cost trend. That answers the conservative's worry that "seasonal" excuses a miss, because the pass depends on Q3 against the prior guide and not on the explanation alone. Write down the Q3 number you'd call a miss before you look.

Here's what I'd keep from this whole exchange. The starter is rounding error: about nine or ten basis points lost at the weekly line, about fifteen made at 434. Nobody here has a measurable edge. The conservative says that means the rules have to carry the plan, and I agree, because it's an argument for staging. With no edge, the only way to get paid for owning a lumpy stock is to keep the first loss small and let size follow price.

The gates that survived do exactly that: the release gates the starter, and price proof gates each add. The ones that died were taxes on the normal path, such as stops inside one day's range, trimming below standard, and waiting on unverified peer dates.

The one path this plan can't fully capture is a clean read followed by a gap-and-go that never lets the bid fill. The breakout half is the price of being in that. It risks five or six basis points, and it opens a runner with a trail to entry and a path toward 453. I'll take that bet every time.

So here's where I land: - Over your ceiling: sell the excess into 398-404 in clips. It goes anyway on a close under 372.80 or after a failed read. - At or below your ceiling: sit, and don't add at 395. - Flat: read the release against the written criteria, including Q3 against the old guide path and the reason behind the implied Q4. A missing or ambiguous read means no starter. Then work a third of your ceiling on day orders at 388-391, placed after the first half hour. Re-arm them each morning, and not during an active peer veto. - Second piece, in two halves: the first after a session trades into 386-388 and closes above 392 with no daily close under 386. The second after a second close over 404 on roughly 6-7M shares. - Last piece: sized by its distance to the failure line, and only once MACD is above zero and ADX is over 25. No adds above your last sale, and no catch-up buying. - Exits: sell at 425 and 434, trail to entry, and run the remainder toward 453. 379.60 is an alert. A close under 372.80 halves everything above the starter. A weekly flip halves the rest.

Final transaction proposal: HOLD.