soda-topic-selection

SODA Topic Selection

SODA's identity: the joint ACM-SIAM venue for the design and analysis of

efficient algorithms and data structures for discrete problems, including the

combinatorics and discrete mathematics that feed them, and admitting

experimental validation alongside theory (SIAM SODA conference pages, checked

2026-07-08). The reviewing bench is the deepest algorithms-specialist pool of

any annual venue — which is precisely the fit test: SODA rewards papers whose

value is legible to algorithms experts as progress on concrete problems, not

papers needing a general theory-of-computation audience or a systems audience.

The three-question fit test

1. Is there a problem with a bound? SODA papers move a quantitative frontier:

running time, space, approximation ratio, query complexity, update time,

list size, competitive ratio. If your contribution cannot be stated as a

before/after on some measure of some problem, the fit is doubtful.

1. Does the improvement need an expert to be appreciated? SODA's specialist

bench is an advantage exactly when the answer is yes — a log shaved off a

bound that experts know is hard earns more at SODA than at a general venue

where "only a log factor" is a slur.

1. Is the technique discrete-algorithmic? Combinatorial, graph-theoretic,

geometric, algebraic, probabilistic — the toolbox of the community. A result

proved entirely with, say, cryptographic machinery may be better read

elsewhere even if it mentions running time.

Two of three is submission territory; the third question then picks the frame.

Routing map from SODA's seat

| Signal in the work | Consider instead | Why |

|---|---|---|

| The point is a new complexity-theoretic barrier or model | STOC / FOCS / CCC | Cross-area significance is their filter; SODA wants the algorithm |

| Bold new problem or model, modest technical depth | ITCS | Conceptual-first reviewing; SODA referees will ask for the bound |

| Solid algorithmic result, European-cycle timing or ESA-community topic | ESA / ICALP | Same community, different calendar seats |

| The contribution is that a known result becomes simple | SOSA (co-located, papers due August 6, 2026 for 2027) | Simplicity is SOSA's explicit mandate; at SODA it reads as non-novel |

| The contribution is implementation and measurement | ALENEX (co-located, due July 20, 2026) | Experiments are the reviewed object there (soda-experiments) |

| Computational geometry core | SoCG | Its own community and calendar, though SODA takes geometry too |

| Distributed / parallel model is the point | PODC / DISC / SPAA | Model-specialist audiences |

| Learning-theoretic sample/regret bounds | COLT / ALT | Statistical-learning reviewer pool |

| Result deserves 60 pages of record, no deadline pressure | SICOMP / JACM / ACM ToA / Algorithmica | Journals archive what conferences announce |

SODA-shaped in practice: verified exemplars

Recent best-paper-level SODA work shows the shape (verified via SIAM epubs and

institutional award announcements; details in resources/exemplars/library.md):

• Solving sparse linear systems faster than matrix multiplication (SODA 2021) —

a named computational barrier crossed for a concrete problem class.

• Deterministic near-linear time minimum cut in weighted graphs (SODA 2024) —

derandomizing a celebrated randomized bound: pure "problem with a bound."

• Breaking the metric voting distortion barrier (SODA 2024) — a stuck constant

in social choice moved by combinatorial technique.

• Improved list size for folded Reed-Solomon codes (SODA 2025) — a parameter

frontier in coding theory, an algorithms-adjacent lane SODA owns.

The pattern: *named problem, stuck frontier, quantitative movement, discrete

technique.* Note what is absent: new models for their own sake, systems

measurements, and philosophy.

Borderline cases, adjudicated

Three recurring gray zones, with the deciding question for each:

• "My result is about a model, proved with algorithms." A new dynamic-graph

adversary model plus matching bounds: if the bounds are the hard part and the

model formalizes existing practice, SODA. If the model is the provocation and

the bounds are demonstrations, ITCS. Deciding question: *would experts cite

this for the theorem or for the definition?*

• "My result is a lower bound." Conditional hardness (SETH/3SUM/APSP-style)

for a concrete problem is fully SODA-shaped — the frontier moves downward

instead of upward, and the comparison table lists upper bounds you now match.

Circuit or communication lower bounds whose interest is complexity-internal

route to CCC/STOC/FOCS. Deciding question: *does the bound speak to

algorithm designers or to complexity theorists?*

• "My result is both simpler and slightly faster." If the speedup stands on

its own, SODA, with simplicity as a bonus paragraph. If the speedup is a log

factor nobody was stuck on and the real value is a two-page proof of a known

theorem, SOSA — where that is the entire point of the venue. Deciding

question: delete the simplicity claim; is what remains publishable?

Anti-fit signals

Re-route if any of these describes the draft:
- The abstract's key sentence has no O(), no ratio, no complexity measure.
- The honest comparison table has no prior row (new problem: consider ITCS)
  or your row doesn't beat any existing row (consider the journal of record).
- The proof's interest is a connection between fields, with the algorithmic
  corollary routine (STOC/FOCS frame the connection better).
- The value is empirical dominance on real inputs (ALENEX).
- The value is that the proof fits on three pages (SOSA).

Framing once fit is established

The same theorem often supports several frames; at SODA, lead with the problem

lineage ("the complexity of X has been open since [Y]"), not the technique

("we introduce a new framework"). Technique-first framing at SODA invites the

question "what does it buy?" — answer it in the title if possible:

bound-carrying titles are quoted; framework-carrying titles are skimmed.

Output format

[Fit verdict] SODA-shaped / borderline / re-route
[Three-question score] <bound? expert-legible? discrete technique?>
[Route] <SODA, or named alternative with the deciding signal>
[Frame] <problem-lineage lead sentence for the abstract>
[Calendar note] <the live deadline this verdict implies (soda-workflow)>