Reading Hortiz:
What Three Trades Reveal
A Five-Pick Inventory
Joe Hortiz walked into the 2026 NFL Draft holding five picks. He walked out with eight, all by Saturday afternoon, after pulling off three separate trades in three consecutive rounds. By my count, those moves netted the Chargers the equivalent of converting a seventh-rounder into a first-rounder in expected on-field value. By the chart most NFL teams negotiate against, two of those three trades were losses.
That gap is the story.
To get there, we revived a draft chart. We’re calling it SCAV (StormCloud AV), and it’s a regression-smoothed projection of the average career Approximate Value produced at each draft slot. Approximate Value is Pro Football Reference’s single-number measure of a player’s seasonal contribution, designed to be comparable across positions and eras. I built the original version of this chart in 2021 as a FanPost over at Bolts From the Blue, working from every pick made between 2009 and 2018. SB Nation has since deleted the FanPost archive, so the original article no longer exists, but the methodology held up well enough that I’m bringing it back. I’ll publish a separate piece updating the underlying dataset and walking through the math. For now, what matters is what SCAV tries to answer that other charts don’t: what does this pick actually return on the field?
When you put SCAV next to the three other charts in common circulation, something becomes clear. The Jimmy Johnson chart, the one most quoted in trade rooms, doesn’t measure on-field value at all. And Hortiz appears to know it.
Before we get to the trades, the pre-draft picture: Hortiz didn’t enter the 2026 draft with seven picks. He entered with five.
The 2026 seventh-rounder was gone before training camp 2024 ever started, sent to Tennessee in the Elijah Molden trade. The original 2026 fifth-rounder (#162) was gone before Halloween 2025, sent to Baltimore as part of the in-season Odafe Oweh deal that turned into one of the most consequential trades the Chargers made all year. Oweh produced 7.5 sacks in 12 regular season games and three more in the AFC Wild Card win over New England, then signed a four-year, $100 million deal with Washington in March. That contract is projected to net the Chargers a third-round compensatory pick in 2027. Hortiz already won the trade. He’s about to win it again.
But that meant the 2026 draft started with a five-pick inventory: one in each of Rounds 1, 2, 3, 4, and 6. By the time the Chargers were on the clock at #22, Hortiz already needed picks. By the time Day 3 ended, he had eight.
Here’s how he got there.
Four Charts, Two Camps
We’re using four charts in this analysis. They split into two camps based on what they measure.
Where all four agree, a trade is unambiguous. Where the behavioral and outcome camps split, the disagreement itself is the finding.
The chart above is the picture. SCAV (gold) and FS (blue) form the top envelope. JJ (red) and RH (green) collapse to the floor by the end of Round 3. The behavioral and outcome charts have different shapes. The behavioral charts crash through the early rounds, then flatten through Day 3. The outcome charts decline more gradually all the way down. That shape mismatch is what Hortiz trades into. When the slope of the behavioral curve doesn’t match the slope of the outcome curve between two specific picks, a trade across that gap creates value on one chart and not the other. The third trade ran in the opposite direction, by design, and we’ll get there.
Zoom in on the first hundred picks and the four charts fan out clearly:
The pick #1 anomaly is real and worth noting. Rich Hill’s chart drops from 100% to 51% by pick #3 because of the QB premium baked into trade history. When teams trade for the top three picks, they’re almost always paying a quarterback tax. SCAV doesn’t see that tax because it measures career production, not transaction price. After the rapid behavioral regression through pick #10, the behavioral charts continue outpacing outcome chart regression until the values become nearly inconsequential. Trading back from Day 1 and Day 2 picks to stockpile early Day 3 picks is a sound strategy for maximizing production out of a draft class.
The Round 2 Slide
The first trade was the cleanest. Hortiz sent pick #55 to New England in exchange for picks #63, #131, and #202. By SCAV, that’s a swap of 22.70 expected AV for 38.06. Hortiz turned a single Round 2 selection into a 67% bump in expected on-field value.
Three of four charts agreed. FS gave the Chargers a +770 point edge. Rich Hill called it +4. SCAV called it +15.36.
Only Jimmy Johnson disagreed, by 23 points. Less than a 7th-rounder’s worth.
This is the cleanest trade-down-for-volume play in the book. The picks acquired weren’t filler. Pick #131 became safety Genesis Smith. Pick #202 became guard Logan Taylor. Two future contributors in a draft where the Chargers needed to fill out the back end of the depth chart at multiple positions. Hortiz traded a single mid-Round 2 pick into two of them.
The most interesting tell is the JJ disagreement. New England’s front office likely walked out of that negotiation believing they got value, because their negotiation tool said they did. By every other measure, they didn’t.
The Round 3 Vacate
The second trade was the loudest. Hortiz sent pick #86 to Cleveland for picks #105, #145, and #206. He didn’t trade down within Round 3. He traded out of it entirely.
By SCAV, the math is staggering: 16.99 expected AV in, 29.70 out. Hortiz nearly doubled the expected value of the pick. This was the biggest absolute SCAV win of the three trades.
Here’s where the chart story gets interesting. JJ said the Chargers gave up 34 points of value. FS said they gained 677. Same trade. Same picks. Different math.
The most useful data point is what Rich Hill said: dead even. 49 to 49. Hill’s chart was specifically built to model what NFL front offices actually pay for picks in the modern CBA era. When Hill says a trade is even, it means the market prices it as fair based on the most recent trading behaviors.
But the picks Hortiz got back outperform that market price by every outcome measure. The story is simple: NFL teams have been undervaluing Day 3, especially early Day 3 picks, relative to their actual expected production for thirty years. Hortiz traded into that gap.
The picks acquired in this deal: #105 became Brenen Thompson, the Mississippi State wideout Hortiz had a late Day 2 grade on. He told reporters in his post-draft presser that Thompson was the player they would have taken at #86 if they couldn’t trade down. #145 became Nick Barrett, the South Carolina defensive lineman who profiles as a rotational run-stopper with flashes of significant upside. #206 became Alex Harkey, an Oregon tackle projected to kick inside.
Three Day 3 picks. Three roster contributors. The cost was a single Round 3 selection in a year where Hortiz had a Day 2 player still on the board after the trade. Hortiz values Day 3 volume more than the behavioral charts price it. That isn’t a hot take. It’s a strategy with an analytical foundation.
The Round 4 Climb
The third trade is where the pattern inverts.
Hortiz sent picks #123 and #204 to Houston in exchange for pick #117. He climbed six spots to lock in Memphis offensive tackle Travis Burke. By SCAV, the trade cost 5.02 expected AV. The Chargers gave up 17.77 in expected on-field value to acquire 12.75.
For the first time, three of the four charts said Hortiz overpaid. FS said -279. RH said -2. SCAV said -5.02.
The only chart that gave the Chargers a win was Jimmy Johnson. By two points. Within rounding error.
Take that in for a moment. In every trade Hortiz made on draft day, the Jimmy Johnson chart told the opposite story from every other chart. Trade 1: JJ said New England won. Trade 2: JJ said Cleveland won. Trade 3: JJ said the Chargers won. SCAV said the exact opposite each time. So did FS. So did RH, mostly.
But before grading this trade-up as a loss, look at what Hortiz has done with this exact same play before.
The Trade-Up Ledger
Burke isn’t the first time Hortiz has paid a small SCAV premium to climb the board. It’s the third.
The pattern across all three trade-ups is consistent. Hortiz pays small premiums to climb. He pays them in late Day 3 capital, either literally or by sliding a mid-round pick down a round. And so far, he has been right about the player he climbed for every time.
The math on the cost is striking. The average SCAV loss across the three trade-ups is 3.81 AV. The average SCAV gain on the two 2026 trade-downs is 14.04 AV. The price Hortiz pays to climb is barely a quarter of what he gains on the way down. That isn’t an accident. It’s a tell. He only moves up if the SCAV loss is marginal and costs no more than one extra late Day 3 pick. That small of a margin is where scouting proficiency will often outproduce projections.
That last part is what makes this work. Trade-up math almost always shows a loss on the FS or SCAV charts, because teams behaviorally overvalue the higher rounds and the slot prices reflect that. The question isn’t whether the chart shows a loss. The question is whether the player outproduces the slot. McConkey and Still have already buried that math. Gadsden looks well on his way. Burke is the next bet.
Burke is 6’9″, 325 pounds, First-Team All-AAC, with a 98.0 PFF pass-blocking efficiency grade in 2025. The traits are there. The bet is on the player, not the slot.
Both Sides of the Asymmetry
Step back from the individual trades and the bigger picture sharpens.
Net SCAV across the three 2026 deals: +23.05. That’s roughly the equivalent of converting pick #200 into pick #25. In a draft where the Chargers entered with a five-pick inventory and needed to fill multiple roster holes, that’s a meaningful win.
But the more interesting finding is structural. Jimmy Johnson disagreed with SCAV on every single 2026 trade. Every one. And in two of the three cases, JJ was the only chart in disagreement.
If the chart NFL teams use to negotiate against Hortiz consistently misvalues his trade-down behavior, then his trade partners may feel like they’re winning on the books that matter to them while losing on the books that matter on Sundays. That’s not an accident. That’s an exploitable inefficiency, and it’s been hiding in plain sight since Belichick’s Patriots started running variations of this same playbook in the early 2000s.
The trade-up math runs in the opposite direction, and Hortiz handles it differently. He doesn’t climb unless the cost fits a specific shape: a piece of late Day 3 capital when he is already drafting from a pick count surplus. Picks in the 6th and 7th round have largely been used by Hortiz as trade chips, both to secure cut-candidate players (Taylor Heinicke, Elijah Molden, and Trevor Penning were all traded for 7s) or for these marginal move-ups. The names on his ledger (McConkey, Still, Gadsden) suggest he’s earned the right to that approach. He pays in pick volume to get conviction picks, but only when the price stays small. That’s the trade he’s making, and it’s the disciplined inverse of the trade-down play. Both can work at the same time, especially when Hortiz regularly manages to stockpile extra picks in his war chest.
The Lane Hortiz Operates In
The league has shifted, with more progressive and analytically driven teams incorporating modern draft charts in war rooms, but trades still fall closer to the JJ and RH values because behaviors haven’t significantly shifted on the aggregate. Especially in franchises where job security isn’t guaranteed year over year, GMs are pressured to make safer, more popular decisions. There aren’t many greater dopamine dumps in the offseason than drafting a Day 1 or Day 2 highly coveted prospect that projects as an immediate starter upgrade.
Capitalizing on Inefficiency, Trusting the Process
What Hortiz is doing is capitalizing on the emotional tie GMs have to higher picks while trusting his scouting process to score gems in the later rounds, as they have consistently done. It’s a market inefficiency that plays into Hortiz’s greatest strength as a GM, and it costs teams whatever the SCAV gap happens to be.
For the Chargers, that gap turned five draft picks into eight contributors, plus a 2027 third-round comp pick already in the books from the Oweh deal that started this whole chain in October.
The capital cycle never stops. It just gets passed forward, one trade at a time. By a guy who appears to know exactly which chart he’s playing on.
SCAV (StormCloud AV Chart) is a regression-smoothed projection of expected career Approximate Value at each draft slot, built from Pro Football Reference data covering the 2009 through 2018 NFL Drafts. The original version of this chart was published as a Bolts From the Blue FanPost in 2021. A full methodology update is forthcoming. All trade values, pick assignments, and player-to-pick mappings verified via Pro Football Reference. SCAV calculations use double-exponential regression on the 5-year tAV column.
Using the Baldwin chart:
Trade 1:
Gave pick 55 = 64 points
Got pick 63 = 56 points
Got pick 131 = 21 points
Got pick 202 = 7 points
Net: traded away 64 points for 84 points. Win.
Trade 2:
Gave pick 86 = 39 points
Got pick 105 = 30 points
Got pick 145 = 19 points
Got pick 206 = 6 points
Net: traded away 39 points for 55 points. Win.
Trade 3:
Gave pick 123 = 24 points
Gave pick 204 = 7 points
Got pick 117 = 26 points
Net: traded away 31 points for 26 points. Loss. Mild but reasonable overpay.
Overall net: traded away 134 points for 165 points while moving from 5 picks to 8 picks. Very strong process, especially given the generally held view that the draft was below average in talent.
Very strong process doesn’t equate to very strong outcomes. We need to see the players selected with those picks perform well on the field. But there is reason for optimism based on their draft profiles and Hortiz’s track record.
This is another fantastic article Kyle. I love the breakdown of the philosophy used during the draft. I think another big factor is that IMO, this draft lacked a ton of star power/blue chip prospects. There were a bunch of players that probably graded out similarly, so they probably felt the value was more selections from this tier of players. While they might not find starters from these selections like previous drafts, they could stock the team with competition and depth. Some fans forget that not all drafts are equal. Next year there might be later round prospects with higher grades that can be targeted for bigger roles, but that didn’t present itself here. While Hortiz has earned the criticism for his FA approach, I don’t think anyone can fault his draft day prowess and prospect evaluation. Those have been top tier IMO.
Agreed, Erick! His drafting is pretty dang impressive. I think the good news is we are only a year or two away from not having a glut of cap space in the offseason, so at least he won’t piss us all off by not spending available cap space on players we are all convinced will turn us into true contenders!
Here is a Ben Baldwin draft value chart excluding QBs: NFL Draft Value Chart (https://opensourcefootball.com/posts/2023-02-23-nfl-draft-value-chart/).
(Note: links are not saving, which is why I pasted in the URL.)
I found this very interesting.
I may have posted about AV previously here, because I know have posted many times about it in various places I post over the past 15 years, but it is flawed.
Here is the first in a series of articles describing AV. I had to go to the wayback machine to find it, as the PFR link is broken. Issues with AV:
1. Assumption 1: the offensive line is exactly as good as the offense. This is obviously flawed, and Doug Drinen, AV creator admits it.
2. Assumption #2 : the offensive line is equally important in the running game as it is in the passing game. Drinen admitted this may be flawed, but he wasn’t convinced of that.
3. Assumption #3 : the ratio of pass-thrower importance to pass-catcher importance is constant from team to team. This is obviously flawed, and Drinen admitted this.
4. Team AV points are based on points per drive scored/allowed. No elements for yards gained/allowed or any other metrics.
5. Among other things, AV does not account for strength of schedule, injuries, coaching impact (think Greg Roman here), or teammate impact (e.g., WRs playing with a terrible passing QB). At least none of these are accounted for directly, only indirectly via impact on Team AV points.
6 Drinen basically admits that he doesn’t know how to positionally divide up OL credit, which is not surprising since that is a complex problem. But his answer is to assign fixed positional values. So a tackle gets 20% more credit than a guard or center. Sometimes that makes sense, but other times it doesn’t.
7. Offensive points are mostly based on yards, ignoring receptions, TDs, first downs, sacks taken, and offensive turnovers. This results in situations like this one for the 2018 Chargers:
Mike obviously had a more positively impactful season than Tyrell, with 6 more TDs, 8 more first downs, 24 more YFS, 1 fewer drop, and 2 fewer fumbles… but equal AV.
8. Credit for efficiency is limited. For example, runners only get an adjustment (up or down) for YPC if they have 200+ carries in a season.
9. For defense:
1/3 of the team points are assigned to defensive backs and 2/3 to the rest of the position groups, independent of how good those groups are on any given defense.
Other defensive issues:
10. AV does not account for postseason games.
I get that AV is an approximation tool, and that’s fine. It was ahead of its time when Doug published it in January 2008. But football analytics have obviously advanced tremendously since then, and AV now looks pretty dated. I expect if someone set out to create a better version of AV today, the results that new model produced would be quite different than AV results.
Doug himself said this about AV:
As much as I admire Doug’s work, AV has flaws that are pretty substantial, which unfortunately undermines the utility of SCAV. Perhaps it would be okay if the flaws applied equally to all players (all positions), but they do not.
I suppose you have to decide
Tau837 if, despite the flaws, AV is closer to being useless or useful as a proxy for value, for the purpose of Kyle’s exercise.
I thought the fact that the same metric (albeit flawed) has been applied across every draft pick might make it closer to being useful, even if the flaws don’t apply equally to all position groups.
I know some draft picks tend to be spent more often on certain position groups (eg, QBs and pass rushers have often been drafted between Picks 1-5 over the years), but the fact that he’s applying AV to the ‘picks’ and not to any ‘players’ should help (eg, Pick 20 over the years would have been spent on many different position groups).
I still found this article (which would’ve taken an age for Kyle to think about and write) pretty interesting. Even if you focus on a different value-based metric (eg, Fitzgerald-Spielberger), seeing that Jimmy Johnson and those metrics produce vastly different outcomes was worth thinking about. I’m glad to see that Hortiz is leaning more into the value-based ones personally.
I agree totally that the article was excellent. I’m not sure what to think of SCAV given how I feel about AV, but that doesn’t affect the non-SCAV content at all.
These are fair critiques and clearly show how deep you’ve gone on AV. Where I’d push back is on the conclusion: heavily aggregated, data-heavy queries are exactly where AV is at its strongest. That’s the use case I’m leaning on here.
A quick note on the broader landscape: there are multiple models built on a second-contract input, including OTC’s FS, the Baldwin chart, and one Josh Q put together that actually translates across player-for-pick trades. So the methodology has company.
What I like about AV is that it’s an approachable, aggregated production figure, and where counting stats don’t exist (linemen, for example) a thoughtful process distributes value. It rewards players for contributing on productive teams, and most of the season-level noise smooths out over a career.
On a season-by-season basis, I’m with you. When we’ve talked about single-season or single-game performance and the merits of PFF, I’ve never said “Player A had an AV of 15 in 2013 and Player B had 13, so Player A had the better year.” That’s not what AV is for. The Mike vs. Tyrell 2018 example is exactly where it breaks down, and Drinen himself essentially said as much: AV is a smoothed version of seasons started plus accolades plus counting stats, not a precision instrument.
Where it earns its keep is the career-aggregation level. Run the Mike Williams example out across his full career and the picture sharpens. His career AV of 43 lands him cleanly between Mike Evans at 105 (a clear hit at the same draft slot) and Darrius Heyward-Bey at 26 (a clear miss, same slot). That’s a useful slot. It tells a real story: a productive contributor who didn’t reach the star-caliber bar his draft position implied. Tyrell’s career AV of 27 also fits the shape of his career: an undrafted receiver with four starting seasons, two shortened ones, and a 1,000-yard year. Year by year, AV mis-priced Tyrell against Mike. Career to career, it didn’t.
The Williams vs. Evans comparison is also why I think SCAV has a place at the table. On the second-contract framework, Williams actually earned a higher APY share than Evans ($20M on a $208.2M cap = 9.61%, versus $16.5M on $177.2M = 9.31%). That’s a clean example of why second-contract valuations don’t necessarily reflect first-contract (or even second-contract) production. Evans went 309-4,579-32 over his first four years. Williams went 158-2,516-20.
As Baldwin points out this is the underlying architecture of his chart, I think it’s a good example that all these charts have flaws that we have to consider:
Storm Norton is another one. His 2021 AV of 8 was real (Slater 14, Linsley 12, Schofield and Feiler both at 7 on a top-tier Herbert offense), but his career AV is 14. That single season was more than half his entire NFL value. Feiler ended at 37, Schofield at 44. The 2021 number flattered Storm. The career rollup put him exactly where the eye test had him: a short-window starter, not a long-haul piece.
That’s the pitch. Volume, accolades, and “did he start” bake in over time, and the season-level noise washes out across a career. It doesn’t cure the structural flaws you listed (the 1/3 DB cap, static OL weights, no postseason credit, no pressure/PD/FF credit) and I’m not pretending it does. But for “where does this player’s career sit relative to peers at the same position and draft slot,” I think AV holds up.
Thank you for sharing Baldwin’s chart though. It’s admittedly done by someone far more capable of statical analysis than me. 😂
I will agree to disagree with you about AV utility. It is seriously flawed, as I outlined, so I don’t agree that any other metric based on AV, like SCAV, is useful.
Kyle… Bro… this is awesome. I remember your BTFB fan post. Now I know Stormcloud needs an analytics engine. We show build something. I’ve been fooling around with Codex and PFF data and a githubs nflverse. It would be statically weird if we didnt build something right lol? Also a quick heads up the website logged me in as an admin. I love Stormcloud but I dont want to see how the sausage is made.
Thank you! I’m totally toying around with as many AI tools as possible right now to see what can be of use to us. I currently have AI agent profiles for Jim, Hortiz, OLeary, and McDaniel and had them run mock draft simulations on my browser 😂
anything you can think of building, I’m all ears!
Btw – you should be listed as a “contributor” but not as an “admin.” You should be able to Add Posts (though using the Start a Storm button is best) which will populate in our community posts section. This means you may have a dashboard when visiting the site but it shouldn’t actually give you site- editing permissions.
if that’s not the case, let me know!
It’s my wordpress account glitching I think it has nothing to do with Stormcloud. When I have some time I’ll report back.
I’m building a system that doesn’t try to predict games straight up — it decides when to trust PFF and when to trust the market. The whole idea is the market is usually right, but not always, and PFF has spots where it actually has an edge. The main driver I’m finding is blocking data. I’m trying to find those exact moments where the system gets confidently wrong, fix only those, and stack small edges over time instead of guessing games blindly.
Simply put, I want to know when to trust the weekly blocking grade from the week before and when to ditch it for the market and historical trends. What I’m doing is looking at the change between weekly measurements and correlating that, instead of doing typical residual modeling. Happy to discuss further