Every CTF Model Now Has Its Own Page

[capellini]

We've added a detail page for every model on the CTF leaderboard. Click a model's name to open it.

The page displays various metrics about the model. Percentile bars compare the model's return, risk, and risk-adjusted metrics with other submissions, with help icons to define each metric and give more context as to where this model stands with respect to the others.

A CAPM panel splits performance into alpha and market exposure, and a radar chart compares the model against the others on whichever metrics you pick. Code and PDF documentation download links are also present.

See MAXSER-PR (+ LightGBM) for an example of one.

Share and compare

Every page has a permanent link, so you can cite a result in a paper or drop it into a job-market packet. Post the link on LinkedIn, Bluesky, or Facebook. A preview card showing the model's name, authors, Sharpe ratio, annualized return, and max drawdown will display in the post:

The model detail page contains share buttons for each of those social networks, plus email.

You can compare up to four models at once on the model comparison page, including benchmarks, with a shared radar chart and a full table of metrics. This comparison compares MAXSER-PR against a nonlinear IPCA model and the Instrumented PCA benchmark. Click the Compare button on a model page to start your comparison.

Get your model on the leaderboard

You can view the model detail pages for existing submissions by clicking on the model's name in the leaderboard. If one of them is yours, please share it. New submissions will get a page once they're scored. We'll email you the link along with the headline results.

Start with the submission rules, CTF dataset, and starter code. When your model is ready, submit it here.

Questions? Open a discussion thread or contact us.

[kellybry]

Beautiful! Do users submit a verbal description of their models? Would be great if so to include these on the page.

…

On Wed, Sep 30, 2026 at 11:01 PM Rob Capellini ***@***.***> wrote: We've added a detail page for every model on the CTF leaderboard <https://jkpfactors.com/ctf/leaderboard>. Click a model's name to open it. The page displays various metrics about the model. Percentile bars compare the model's return, risk, and risk-adjusted metrics with other submissions, with help icons to define each metric and give more context as to where this model stands with respect to the others. A CAPM panel splits performance into alpha and market exposure, and a radar chart compares the model against the others on whichever metrics you pick. Code and PDF documentation download links are also present. See MAXSER-PR (+ LightGBM) <https://jkpfactors.com/ctf/models/maxser-pr-lightgbm-x1tz95peaw> for an example of one. Share and compare Every page has a permanent link, so you can cite a result in a paper or drop it into a job-market packet. Post the link on LinkedIn, Bluesky, or Facebook. A preview card showing the model's name, authors, Sharpe ratio, annualized return, and max drawdown will display in the post: [image: Share preview card for MAXSER-PR (+ LightGBM)] <https://jkpfactors.com/ctf/models/maxser-pr-lightgbm-x1tz95peaw> The model detail page contains share buttons for each of those social networks, plus email. You can compare up to four models at once on the model comparison page, including benchmarks, with a shared radar chart and a full table of metrics. This comparison <https://jkpfactors.com/ctf/compare?models=maxser-pr-lightgbm-x1tz95peaw,nonlinear-ipca-with-precision-weighting-cx717n7sdb,benchmark-instrumented-pca-factor-portfolio> compares MAXSER-PR against a nonlinear IPCA model and the Instrumented PCA benchmark. Click the Compare button on a model page to start your comparison. Get your model on the leaderboard You can view the model detail pages for existing submissions by clicking on the model's name in the leaderboard. If one of them is yours, please share it. New submissions will get a page once they're scored. We'll email you the link along with the headline results. Start with the submission rules <https://jkpfactors.com/ctf/rules>, CTF dataset <https://jkpfactors.com/ctf/dataset-access>, and starter code <https://github.com/theisij/common-task-framework-SDF>. When your model is ready, submit it here <https://jkpfactors.com/ctf/submit>. Questions? Open a discussion thread <https://github.com/bkelly-lab/jkp-data/discussions> or contact us <https://jkpfactors.com/contact>. — Reply to this email directly, view it on GitHub <#282?email_source=notifications&email_token=ACICUE4YOAEOVAQTUF23YCT5RXCHHA5CNFSNUABBM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63RPGEYDSMRUHE3DRJTSMVQXG33OVJZXKYTTMNZGSYTFMSSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACICUE2TAVRELUJXVZH6JBL5RXCHHAVCNFSNUABIKJSXA33TNF2G64TZHM3DQNRRHA2TQOJZHNCGS43DOVZXG2LPNY5TCMBZGI2DSNRYUF3AE> . Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/ACICUE2ZTRP7MWSQF2C7AOT5RXCHHA5CNFSNUABBM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63RPGEYDSMRUHE3DRJTSMVQXG33OVJZXKYTTMNZGSYTFMSSWK5TFNZ2KUZTPN52GK4S7NFXXG> and Android <https://github.com/notifications/mobile/android/ACICUE6UECQD6AGG3ZHZCPD5RXCHHA5CNFSNUABBM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63RPGEYDSMRUHE3DRJTSMVQXG33OVJZXKYTTMNZGSYTFMSSWK5TFNZ2K4ZTPN52GK4S7MFXGI4TPNFSA>. Download it today! You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[capellini]

They can optionally include PDF documentation, but currently they aren't required to include something like an abstract summary of their models. Perhaps we can try to use an LLM to create a quick summary of each model? Or would we want this to come directly from the submitters?

[kellybry]

An LLM description is a great idea...it is a consistent mapping from code to text without scope for researcher editorializing. If the LLM gives a decent description, that would be great! Bryan T. Kelly Frederick Frank ’54 and Mary C. Tanner Professor of Finance Yale School of Management Overview of Financial Machine Learning <https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4501707> https://www.bryankellyacademic.org/ https://jkpfactors.com/

…

On Thu, Oct 1, 2026 at 9:54 PM Rob Capellini ***@***.***> wrote: They can optionally include PDF documentation, but currently they aren't required to include something like an abstract summary of their models. Perhaps we can try to use an LLM to create a quick summary of each model? Or would we want this to come directly from the submitters? — Reply to this email directly, view it on GitHub <#282?email_source=notifications&email_token=ACICUE4ABZLIG7Y7EU3UZPT5R4DF3A5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCOBXGA3TONRXUZZGKYLTN5XKOY3PNVWWK3TUUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#discussioncomment-18707767>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACICUE2BXNYANA7BMQFQWQD5R4DF3AVCNFSNUABIKJSXA33TNF2G64TZHM3DQNRRHA2TQOJZHNCGS43DOVZXG2LPNY5TCMBZGI2DSNRYUF3AE> . Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS <https://github.com/notifications/mobile/ios/ACICUE5572G6G6CS5JEJHDD5R4DF3A5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCOBXGA3TONRXUZZGKYLTN5XKOY3PNVWWK3TUUVSXMZLOOSVGM33PORSXEX3JN5ZQ> and Android <https://github.com/notifications/mobile/android/ACICUE7LHKJKQ2S7FSELPIT5R4DF3A5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCOBXGA3TONRXUZZGKYLTN5XKOY3PNVWWK3TUUVSXMZLOOSXGM33PORSXEX3BNZSHE33JMQ>. Download it today! You are receiving this because you commented.Message ID: ***@***.***>